title
string
paper_url
string
authors
list
type
string
primary_area
string
abstract
large_string
keywords
list
TL;DR
large_string
submission_number
int64
arxiv_id
string
arxiv_id_source
string
embedding
list
Multi-Scale Context Aggregation by Dilated Convolutions
https://arxiv.org/abs/1511.07122
[ "Fisher Yu", "Vladlen Koltun" ]
null
null
State-of-the-art models for semantic segmentation are based on adaptations of convolutional networks that had originally been designed for image classification. However, dense prediction and image classification are structurally different. In this work, we develop a new convolutional network module that is specifical...
[]
null
1
1511.07122
iclr_archive
[ 0.002269756980240345, -0.009875677525997162, 0.009501204825937748, 0.01731916330754757, 0.017548799514770508, 0.028615040704607964, 0.015045782551169395, 0.026375215500593185, -0.04363730549812317, -0.026844767853617668, -0.025632137432694435, -0.002348523586988449, -0.05309893935918808, 0...
The Variational Fair Autoencoder
https://arxiv.org/abs/1511.00830
[ "Christos Louizos", "Kevin Swersky", "Yujia Li", "Max Welling", "Richard Zemel" ]
null
null
We investigate the problem of learning representations that are invariant to certain nuisance or sensitive factors of variation in the data while retaining as much of the remaining information as possible. Our model is based on a variational autoencoding architecture with priors that encourage independence between se...
[]
null
2
1511.00830
iclr_archive
[ 0.010011797770857811, 0.010490147396922112, -0.0268756914883852, 0.06136319413781166, 0.02550787851214409, 0.056464362889528275, 0.036562055349349976, 0.004290788900107145, -0.04380892589688301, -0.026354890316724777, -0.009691717103123665, -0.00035323045449331403, -0.08491644263267517, -0...
A note on the evaluation of generative models
https://arxiv.org/abs/1511.01844
[ "Lucas Theis", "Aäron van den Oord", "Matthias Bethge" ]
null
null
Probabilistic generative models can be used for compression, denoising, inpainting, texture synthesis, semi-supervised learning, unsupervised feature learning, and other tasks. Given this wide range of applications, it is not surprising that a lot of heterogeneity exists in the way these models are formulated, traine...
[]
null
3
1511.01844
iclr_archive
[ 0.009832056239247322, -0.02303270250558853, -0.022962285205721855, 0.05108632519841194, 0.05392630398273468, 0.02817285805940628, 0.027782440185546875, 0.007335657719522715, -0.007767823990434408, -0.08241309970617294, -0.024786558002233505, -0.025413045659661293, -0.0635199174284935, -0.0...
Learning to Diagnose with LSTM Recurrent Neural Networks
https://arxiv.org/abs/1511.03677
[ "Zachary Lipton", "David Kale", "Charles Elkan", "Randall Wetzel" ]
null
null
Clinical medical data, especially in the intensive care unit (ICU), consist of multivariate time series of observations. For each patient visit (or episode), sensor data and lab test results are recorded in the patient's Electronic Health Record (EHR). While potentially containing a wealth of insights, the data is di...
[]
null
4
1511.03677
iclr_archive
[ -0.0034203375689685345, -0.016031181439757347, -0.02002669684588909, 0.02639067731797695, 0.06008860096335411, 0.04551081731915474, 0.0466432049870491, 0.025491008535027504, -0.002970160683616996, -0.04646531119942665, 0.009747610427439213, 0.004410638008266687, -0.055326126515865326, 0.01...
Prioritized Experience Replay
https://arxiv.org/abs/1511.05952
[ "Tom Schaul", "John Quan", "Ioannis Antonoglou", "David Silver" ]
null
null
Experience replay lets online reinforcement learning agents remember and reuse experiences from the past. In prior work, experience transitions were uniformly sampled from a replay memory. However, this approach simply replays transitions at the same frequency that they were originally experienced, regardless of thei...
[]
null
5
1511.05952
iclr_archive
[ -0.05309019237756729, -0.01434791088104248, -0.01387566328048706, 0.044200699776411057, 0.05368458852171898, 0.009682840667665005, -0.005636653397232294, 0.0218200646340847, -0.05182097479701042, -0.035653021186590195, -0.013318915851414204, 0.01723213866353035, -0.03419813886284828, -0.02...
Importance Weighted Autoencoders
https://arxiv.org/abs/1509.00519
[ "Yuri Burda", "Ruslan Salakhutdinov", "Roger Grosse" ]
null
null
The variational autoencoder (VAE; Kingma, Welling (2014)) is a recently proposed generative model pairing a top-down generative network with a bottom-up recognition network which approximates posterior inference. It typically makes strong assumptions about posterior inference, for instance that the posterior distribu...
[]
null
6
1509.00519
iclr_archive
[ 0.006149292923510075, 0.0037461097817867994, 0.0025713639333844185, 0.04888444021344185, 0.002296862192451954, 0.0697835311293602, 0.04267501085996628, -0.011665944941341877, -0.022235015407204628, -0.044735781848430634, -0.023736661300063133, 0.004813610576093197, -0.058230478316545486, -...
Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding
https://arxiv.org/abs/1510.00149
[ "Song Han", "Huizi Mao", "Bill Dally" ]
null
null
Neural networks are both computationally intensive and memory intensive, making them difficult to deploy on embedded systems with limited hardware resources. To address this limitation, we introduce "deep compression", a three stage pipeline: pruning, trained quantization and Huffman coding, that work together to red...
[]
null
7
1510.00149
iclr_archive
[ -0.021032879129052162, -0.036423809826374054, -0.050461892038583755, 0.04539690166711807, 0.06357482820749283, 0.07914907485246658, -0.01139076054096222, 0.008869264274835587, -0.015319932252168655, -0.06012798473238945, -0.008755888789892197, -0.014216589741408825, -0.05006549507379532, -...
Variationally Auto-Encoded Deep Gaussian Processes
https://arxiv.org/abs/1511.06455
[ "Zhenwen Dai", "Andreas Damianou", "Javier Gonzalez", "Neil Lawrence" ]
null
null
We develop a scalable deep non-parametric generative model by augmenting deep Gaussian processes with a recognition model. Inference is performed in a novel scalable variational framework where the variational posterior distributions are reparametrized through a multilayer perceptron. The key aspect of this reformula...
[]
null
8
1511.06455
iclr_archive
[ 0.004447946324944496, -0.01196215394884348, -0.010274939239025116, 0.04261202737689018, 0.03276054188609123, 0.045771073549985886, 0.014781394973397255, 0.005644228775054216, -0.016700686886906624, -0.04192059114575386, -0.03154046833515167, -0.0014360854402184486, -0.06329268962144852, 0....
Training Convolutional Neural Networks with Low-rank Filters for Efficient Image Classification
https://arxiv.org/abs/1511.06744
[ "Yani Ioannou", "Duncan Robertson", "Jamie Shotton", "roberto Cipolla", "Antonio Criminisi", "Jamie Shotton" ]
null
null
We propose a new method for creating computationally efficient convolutional neural networks (CNNs) by using low-rank representations of convolutional filters. Rather than approximating filters in previously-trained networks with more efficient versions, we learn a set of small basis filters from scratch; during trai...
[]
null
9
1511.06744
iclr_archive
[ 0.01612483523786068, -0.035694144666194916, 0.03353550285100937, 0.051848411560058594, 0.020053379237651825, 0.026492958888411522, 0.006611557677388191, -0.015075348317623138, -0.02156919799745083, -0.052445344626903534, -0.02003875933587551, -0.0017269409727305174, -0.08760741353034973, 0...
Neural Networks with Few Multiplications
https://arxiv.org/abs/1510.03009
[ "Zhouhan Lin", "Matthieu Courbariaux", "Roland Memisevic", "Yoshua Bengio" ]
null
null
For most deep learning algorithms training is notoriously time consuming. Since most of the computation in training neural networks is typically spent on floating point multiplications, we investigate an approach to training that eliminates the need for most of these. Our method consists of two parts: First we stocha...
[]
null
10
1510.03009
iclr_archive
[ 0.017456823959946632, -0.011907187290489674, -0.00955017376691103, 0.04077688232064247, 0.033862318843603134, 0.040906473994255066, 0.009500469081103802, 0.00433991476893425, -0.030468037351965904, -0.023687118664383888, 0.021049121394753456, -0.0069366442039608955, -0.05132779851555824, 0...
Reducing Overfitting in Deep Networks by Decorrelating Representations
https://arxiv.org/abs/1511.06068
[ "Michael Cogswell", "Faruk Ahmed", "Ross Girshick", "Larry Zitnick", "Dhruv Batra" ]
null
null
One major challenge in training Deep Neural Networks is preventing overfitting. Many techniques such as data augmentation and novel regularizers such as Dropout have been proposed to prevent overfitting without requiring a massive amount of training data. In this work, we propose a new regularizer called DeCov which ...
[]
null
11
1511.06068
iclr_archive
[ 0.016065198928117752, -0.029277630150318146, -0.03194137662649155, 0.023221613839268684, 0.045761529356241226, 0.04238070920109749, 0.03004755638539791, -0.001295856200158596, -0.014115300960838795, -0.04225233942270279, -0.008000941015779972, 0.02174757979810238, -0.0779053270816803, 0.01...
Pushing the Boundaries of Boundary Detection using Deep Learning
https://arxiv.org/abs/1511.07386
[ "Iasonas Kokkinos" ]
null
null
In this work we show that adapting Deep Convolutional Neural Network training to the task of boundary detection can result in substantial improvements over the current state-of-the-art in boundary detection. Our contributions consist firstly in combining a careful design of the loss for boundary detection training,...
[]
null
12
1511.07386
iclr_archive
[ 0.004830592777580023, -0.027814816683530807, 0.010055788792669773, 0.021233122795820236, 0.02936537005007267, 0.0043038115836679935, 0.024321744218468666, -0.000007982550414453726, -0.021398909389972687, -0.06230238825082779, -0.04469341039657593, -0.0011062315898016095, -0.04367852956056595...
Generating Images from Captions with Attention
https://arxiv.org/abs/1511.02793
[ "Elman Mansimov", "Emilio Parisotto", "Jimmy Ba", "Ruslan Salakhutdinov" ]
null
null
Motivated by the recent progress in generative models, we introduce a model that generates images from natural language descriptions. The proposed model iteratively draws patches on a canvas, while attending to the relevant words in the description. After training on Microsoft COCO, we compare our model with several ...
[]
null
13
1511.02793
iclr_archive
[ -0.005113136488944292, -0.023145699873566628, -0.02082972787320614, 0.07663147896528244, 0.018557894974946976, 0.0112132104113698, -0.002434214809909463, 0.052130475640296936, -0.020421423017978668, -0.02231953851878643, -0.06719367206096649, 0.0134742958471179, -0.05635499581694603, 0.013...
Reasoning about Entailment with Neural Attention
https://arxiv.org/abs/1509.06664
[ "Tim Rocktäschel", "Edward Grefenstette", "Karl Moritz Hermann", "Tomáš Kočiský", "Phil Blunsom" ]
null
null
While most approaches to automatically recognizing entailment relations have used classifiers employing hand engineered features derived from complex natural language processing pipelines, in practice their performance has been only slightly better than bag-of-word pair classifiers using only lexical similarity. The ...
[]
null
14
1509.06664
iclr_archive
[ -0.012275771237909794, -0.009826903231441975, -0.006962043233215809, 0.041838161647319794, 0.03753141686320305, 0.022171173244714737, 0.01774180307984352, 0.009482880122959614, -0.028990818187594414, 0.02131958119571209, -0.0428452305495739, 0.07207783311605453, -0.03173317387700081, 0.003...
Convolutional Neural Networks With Low-rank Regularization
https://arxiv.org/abs/1511.06067
[ "Cheng Tai", "Tong Xiao", "Yi Zhang", "Xiaogang Wang", "Weinan E" ]
null
null
Large CNNs have delivered impressive performance in various computer vision applications. But the storage and computation requirements make it problematic for deploying these models on mobile devices. Recently, tensor decompositions have been used for speeding up CNNs. In this paper, we further develop the tensor dec...
[]
null
15
1511.06067
iclr_archive
[ -0.027476755902171135, -0.03488822281360626, 0.02248748019337654, 0.03587082028388977, 0.02984938770532608, 0.0269013624638319, -0.00539483642205596, -0.020633075386285782, -0.03365094214677811, -0.04408653825521469, -0.0030361246317625046, 0.002694296883419156, -0.05975145101547241, 0.031...
Unifying distillation and privileged information
https://arxiv.org/abs/1511.03643
[ "David Lopez-Paz", "Leon Bottou", "Bernhard Schölkopf", "Vladimir Vapnik" ]
null
null
Distillation (Hinton et al., 2015) and privileged information (Vapnik & Izmailov, 2015) are two techniques that enable machines to learn from other machines. This paper unifies these two techniques into generalized distillation, a framework to learn from multiple machines and data representations. We provide theoreti...
[]
null
16
1511.03643
iclr_archive
[ -0.011432897299528122, -0.01155119389295578, -0.022094978019595146, 0.06010114401578903, 0.062361013144254684, -0.0036348996218293905, 0.03660079836845398, -0.021718066185712814, -0.002492764266207814, -0.03574894741177559, -0.024499008432030678, 0.0013737689005210996, -0.06905361264944077, ...
Particular object retrieval with integral max-pooling of CNN activations
https://arxiv.org/abs/1511.05879
[ "[code] Giorgos Tolias", "Ronan Sicre", "Hervé Jégou" ]
null
null
Recently, image representation built upon Convolutional Neural Network (CNN) has been shown to provide effective descriptors for image search, outperforming pre-CNN features as short-vector representations. Yet such models are not compatible with geometry-aware re-ranking methods and still outperformed, on some parti...
[]
null
17
1511.05879
iclr_archive
[ -0.004633492324501276, -0.03312151879072189, -0.014093857258558273, 0.051418814808130264, 0.026407312601804733, 0.021889299154281616, -0.008791995234787464, 0.025567280128598213, -0.009125792421400547, -0.05804632976651192, -0.04571717232465744, -0.028623618185520172, -0.045564815402030945, ...
All you need is a good init
https://arxiv.org/abs/1511.06422
[ "[code] Dmytro Mishkin", "Jiri Matas" ]
null
null
Layer-sequential unit-variance (LSUV) initialization - a simple method for weight initialization for deep net learning - is proposed. The method consists of the two steps. First, pre-initialize weights of each convolution or inner-product layer with orthonormal matrices. Second, proceed from the first to the final la...
[]
null
18
1511.06422
iclr_archive
[ 0.016672834753990173, -0.03189375624060631, -0.015295136719942093, 0.004228948149830103, 0.06121864914894104, 0.041103545576334, 0.008022530935704708, -0.0068925912491977215, -0.01663791574537754, -0.05255363881587982, 0.0071788146160542965, -0.004415791016072035, -0.0570450983941555, -0.0...
Bayesian Representation Learning with Oracle Constraints
https://arxiv.org/abs/1506.05011
[ "Theofanis Karaletsos", "Serge Belongie", "Gunnar Rätsch" ]
null
null
Representation learning systems typically rely on massive amounts of labeled data in order to be trained to high accuracy. Recently, high-dimensional parametric models like neural networks have succeeded in building rich representations using either compressive, reconstructive or supervised criteria. However, the sem...
[]
null
19
1506.05011
iclr_archive
[ 0.006664684973657131, 0.026571298018097878, -0.03676401823759079, 0.04683399945497513, 0.030480267480015755, 0.027709735557436943, 0.00458972854539752, 0.007781757041811943, -0.0282566100358963, -0.021681005135178566, -0.03435336425900459, 0.010743233375251293, -0.07016624510288239, -0.006...
Neural Programmer: Inducing Latent Programs with Gradient Descent
https://arxiv.org/abs/1511.04834
[ "Arvind Neelakantan", "Quoc Le", "Ilya Sutskever" ]
null
null
Deep neural networks have achieved impressive supervised classification performance in many tasks including image recognition, speech recognition, and sequence to sequence learning. However, this success has not been translated to applications like question answering that may involve complex arithmetic and logic reas...
[]
null
20
1511.04834
iclr_archive
[ -0.004740984179079533, -0.012547033838927746, -0.03202664852142334, 0.031895436346530914, 0.048139844089746475, 0.045360565185546875, 0.0071688019670546055, 0.0025399322621524334, -0.03716539964079857, -0.023163899779319763, -0.002564756665378809, 0.026013750582933426, -0.04519571736454964, ...
Towards Universal Paraphrastic Sentence Embeddings
https://arxiv.org/abs/1511.08198
[ "[code] John Wieting", "Mohit Bansal", "Kevin Gimpel", "Karen Livescu" ]
null
null
We consider the problem of learning general-purpose, paraphrastic sentence embeddings based on supervision from the Paraphrase Database (Ganitkevitch et al., 2013). We compare six compositional architectures, evaluating them on annotated textual similarity datasets drawn both from the same distribution as the trainin...
[]
null
21
1511.08198
iclr_archive
[ 0.02366277016699314, -0.0320424884557724, -0.0024995317216962576, 0.0678134337067604, 0.029692796990275383, 0.045228417962789536, 0.030741915106773376, 0.019432369619607925, 0.010072865523397923, -0.007313068490475416, -0.029373489320278168, -0.0007848837994970381, -0.06300512701272964, -0...
Regularizing RNNs by Stabilizing Activations
https://arxiv.org/abs/1511.08400
[ "David Krueger", "Roland Memisevic" ]
null
null
We stabilize the activations of Recurrent Neural Networks (RNNs) by penalizing the squared distance between successive hidden states' norms. This penalty term is an effective regularizer for RNNs including LSTMs and IRNNs, improving performance on character-level language modeling and phoneme recognition, and outpe...
[]
null
22
1511.08400
iclr_archive
[ -0.008789228275418282, -0.03423594683408737, -0.011286403983831406, 0.02845042198896408, 0.05928419902920723, 0.06148194149136543, 0.06261321157217026, -0.00011336096213199198, -0.08411083370447159, -0.03164774179458618, 0.014506439678370953, 0.00730540556833148, -0.05391409620642662, -0.0...
SparkNet: Training Deep Networks in Spark
https://arxiv.org/abs/1511.06051
[ "Philipp Moritz", "Robert Nishihara", "Ion Stoica", "Michael Jordan" ]
null
null
Training deep networks is a time-consuming process, with networks for object recognition often requiring multiple days to train. For this reason, leveraging the resources of a cluster to speed up training is an important area of work. However, widely-popular batch-processing computational frameworks like MapReduce an...
[]
null
23
1511.06051
iclr_archive
[ -0.008420904166996479, -0.06021135672926903, -0.030379772186279297, 0.06683852523565292, 0.033998847007751465, 0.02302977815270424, -0.002671829890459776, 0.009307974018156528, -0.017028043046593666, -0.054326076060533524, -0.003529365174472332, -0.023373639211058617, -0.053127288818359375, ...
Unsupervised and Semi-supervised Learning with Categorical Generative Adversarial Networks
https://arxiv.org/abs/1511.06390
[ "Jost Tobias Springenberg" ]
null
null
In this paper we present a method for learning a discriminative classifier from unlabeled or partially labeled data. Our approach is based on an objective function that trades-off mutual information between observed examples and their predicted categorical class distribution, against robustness of the classifier to a...
[]
null
24
1511.06390
iclr_archive
[ 0.014457511715590954, -0.0388360396027565, -0.02026732824742794, 0.03532547876238823, 0.029661988839507103, 0.006024224683642387, 0.011556042358279228, -0.014774140901863575, -0.01676820032298565, -0.035555873066186905, -0.029545707628130913, 0.00792231597006321, -0.06682967394590378, 0.02...
The Goldilocks Principle: Reading Children's Books with Explicit Memory Representations
https://arxiv.org/abs/1511.02301
[ "Felix Hill", "Antoine Bordes", "Sumit Chopra", "Jason Weston" ]
null
null
We introduce a new test of how well language models capture meaning in children's books. Unlike standard language modelling benchmarks, it distinguishes the task of predicting syntactic function words from that of predicting lower-frequency words, which carry greater semantic content. We compare a range of state-of-t...
[]
null
25
1511.02301
iclr_archive
[ -0.055827487260103226, -0.00605232547968626, -0.0035390870179980993, 0.02132115699350834, 0.040419936180114746, 0.00460384925827384, 0.029065025970339775, 0.028863439336419106, -0.02467224933207035, 0.0066641950979828835, -0.020982516929507256, 0.013029864057898521, -0.04580703750252724, 0...
MuProp: Unbiased Backpropagation For Stochastic Neural Networks
https://arxiv.org/abs/1511.05176
[ "Shixiang Gu", "Sergey Levine", "Ilya Sutskever", "Andriy Mnih" ]
null
null
Deep neural networks are powerful parametric models that can be trained efficiently using the backpropagation algorithm. Stochastic neural networks combine the power of large parametric functions with that of graphical models, which makes it possible to learn very complex distributions. However, as backpropagation is...
[]
null
26
1511.05176
iclr_archive
[ -0.008721927180886269, -0.02783004194498062, -0.01094829011708498, 0.035408370196819305, 0.045895978808403015, 0.04191083833575249, 0.029016030952334404, -0.0030195279978215694, -0.02248000167310238, -0.054388225078582764, 0.05542580038309097, 0.01769467629492283, -0.07341285794973373, -0....
Data Representation and Compression Using Linear-Programming Approximations
https://arxiv.org/abs/1511.06606
[ "Hristo Paskov", "John Mitchell", "Trevor Hastie" ]
null
null
We propose `Dracula', a new framework for unsupervised feature selection from sequential data such as text. Dracula learns a dictionary of $n$-grams that efficiently compresses a given corpus and recursively compresses its own dictionary; in effect, Dracula is a `deep' extension of Compressive Feature Learning. It re...
[]
null
27
1511.06606
iclr_archive
[ -0.020966369658708572, -0.023351656273007393, -0.02360581047832966, 0.03823109343647957, 0.052906718105077744, 0.0334806889295578, 0.017877226695418358, 0.00030389497987926006, -0.012009549885988235, -0.02544383704662323, -0.030158309265971184, 0.023588262498378754, -0.06983702629804611, -...
Diversity Networks
https://arxiv.org/abs/1511.05077
[ "Zelda Mariet", "Suvrit Sra" ]
null
null
We introduce Divnet, a flexible technique for learning networks with diverse neurons. Divnet models neuronal diversity by placing a Determinantal Point Process (DPP) over neurons in a given layer. It uses this DPP to select a subset of diverse neurons and subsequently fuses the redundant neurons into the selected one...
[]
null
28
1511.05077
iclr_archive
[ 0.005854323040693998, -0.028403719887137413, -0.031006639823317528, 0.06313785165548325, 0.029269881546497345, 0.03457092121243477, 0.015883080661296844, 0.0014129607006907463, -0.04921763762831688, -0.04364277422428131, 0.010105142369866371, -0.02167452871799469, -0.06796605885028839, -0....
Deep Reinforcement Learning in Parameterized Action Space
https://arxiv.org/abs/1511.04143
[ "[code] [data] Matthew Hausknecht", "Peter Stone" ]
null
null
Recent work has shown that deep neural networks are capable of approximating both value functions and policies in reinforcement learning domains featuring continuous state and action spaces. However, to the best of our knowledge no previous work has succeeded at using deep neural networks in structured (parameterized...
[]
null
29
1511.04143
iclr_archive
[ -0.03070882335305214, -0.037217941135168076, -0.03566088154911995, 0.04195002466440201, 0.04413206875324249, 0.02735978551208973, 0.031607482582330704, -0.009272860363125801, -0.0016665845178067684, -0.045565228909254074, 0.014955976977944374, -0.00841462705284357, -0.0871310904622078, -0....
Learning VIsual Predictive Models of Physics for Playing Billiards
https://arxiv.org/abs/1511.07404
[ "Katerina Fragkiadaki", "Pulkit Agrawal", "Sergey Levine", "Jitendra Malik" ]
null
null
The ability to plan and execute goal specific actions in varied, unexpected settings is a central requirement of intelligent agents. In this paper, we explore how an agent can be equipped with an internal model of the dynamics of the external world, and how it can use this model to plan novel actions by running multi...
[]
null
30
1511.07404
iclr_archive
[ -0.03584618121385574, 0.007647608872503042, -0.006646361202001572, 0.013518614694476128, 0.04665585234761238, -0.00015270194853655994, 0.008561656810343266, 0.017230715602636337, -0.06477173417806625, -0.04469435662031174, -0.03070227988064289, 0.008065261878073215, -0.06438780575990677, -...
Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks
https://arxiv.org/abs/1502.05698
[ "[code] [data] Jason Weston", "Antoine Bordes", "Sumit Chopra", "Sasha Rush", "Bart van Merrienboer", "Armand Joulin", "Tomas Mikolov" ]
null
null
One long-term goal of machine learning research is to produce methods that are applicable to reasoning and natural language, in particular building an intelligent dialogue agent. To measure progress towards that goal, we argue for the usefulness of a set of proxy tasks that evaluate reading comprehension via question...
[]
null
31
1502.05698
iclr_archive
[ -0.007688414771109819, -0.019020214676856995, -0.010610266588628292, 0.05404873937368393, 0.047384925186634064, 0.018264370039105415, 0.011103159748017788, 0.013258121907711029, -0.03059338964521885, 0.008192448876798153, -0.037784963846206665, 0.07557123154401779, -0.05022421479225159, -0...
Evaluating Prerequisite Qualities for Learning End-to-end Dialog Systems
https://arxiv.org/abs/1511.06931
[ "[data] Jesse Dodge", "Andreea Gane", "Xiang Zhang", "Antoine Bordes", "Sumit Chopra", "Alexander Miller", "Arthur Szlam", "Jason Weston" ]
null
null
A long-term goal of machine learning is to build intelligent conversational agents. One recent popular approach is to train end-to-end models on a large amount of real dialog transcripts between humans (Sordoni et al., 2015; Vinyals & Le, 2015; Shang et al., 2015). However, this approach leaves many questions unanswe...
[]
null
32
1511.06931
iclr_archive
[ -0.0009041887824423611, -0.019766703248023987, -0.007086369674652815, 0.06882129609584808, 0.047691456973552704, 0.01752314902842045, 0.011031007394194603, 0.016064396128058434, 0.016307909041643143, -0.005938160698860884, -0.04745813459157944, 0.08708975464105606, -0.054042380303144455, -...
Better Computer Go Player with Neural Network and Long-term Prediction
https://arxiv.org/abs/1511.06410
[ "Yuandong Tian", "Yan Zhu" ]
null
null
Competing with top human players in the ancient game of Go has been a long-term goal of artificial intelligence. Go's high branching factor makes traditional search techniques ineffective, even on leading-edge hardware, and Go's evaluation function could change drastically with one stone change. Recent works [Maddiso...
[]
null
33
1511.06410
iclr_archive
[ -0.036803536117076874, -0.05119098350405693, -0.0007526528206653893, 0.029567258432507515, 0.06489233672618866, 0.014193223789334297, 0.029702145606279373, 0.033093951642513275, -0.032912541180849075, -0.051505934447050095, -0.0018145284848287702, -0.012201289646327496, -0.06886669248342514,...
Distributional Smoothing with Virtual Adversarial Training
https://arxiv.org/abs/1507.00677
[ "[code] Takeru Miyato", "Shin-ichi Maeda", "Masanori Koyama", "Ken Nakae", "Shin Ishii" ]
null
null
We propose local distributional smoothness (LDS), a new notion of smoothness for statistical model that can be used as a regularization term to promote the smoothness of the model distribution. We named the LDS based regularization as virtual adversarial training (VAT). The LDS of a model at an input datapoint is def...
[]
null
34
1507.00677
iclr_archive
[ 0.02795182541012764, -0.02129034325480461, 0.0022770126815885305, 0.06458956748247147, 0.030864311382174492, 0.02289973758161068, 0.02651953510940075, -0.036681074649095535, -0.017257681116461754, -0.05134053900837898, -0.04053223133087158, -0.009261101484298706, -0.05653007701039314, 0.01...
Multi-task Sequence to Sequence Learning
https://arxiv.org/abs/1511.06114
[ "Minh-Thang Luong", "Quoc Le", "Ilya Sutskever", "Oriol Vinyals", "Lukasz Kaiser" ]
null
null
Sequence to sequence learning has recently emerged as a new paradigm in supervised learning. To date, most of its applications focused on only one task and not much work explored this framework for multiple tasks. This paper examines three multi-task learning (MTL) settings for sequence to sequence models: (a) the on...
[]
null
35
1511.06114
iclr_archive
[ 0.03358972817659378, -0.029909681528806686, -0.01631055399775505, 0.020978108048439026, 0.03315417096018791, 0.014755801297724247, 0.0426291860640049, 0.02693195641040802, 0.00760682625696063, -0.04101351276040077, -0.012809453532099724, 0.029355913400650024, -0.06078207865357399, -0.01256...
A Test of Relative Similarity for Model Selection in Generative Models
https://arxiv.org/abs/1511.04581
[ "Eugene Belilovsky", "Wacha Bounliphone", "Matthew Blaschko", "Ioannis Antonoglou", "Arthur Gretton" ]
null
null
Probabilistic generative models provide a powerful framework for representing data that avoids the expense of manual annotation typically needed by discriminative approaches. Model selection in this generative setting can be challenging, however, particularly when likelihoods are not easily accessible. To address thi...
[]
null
36
1511.04581
iclr_archive
[ -0.010373447090387344, -0.019181372597813606, -0.011309478431940079, 0.06843285262584686, 0.050808750092983246, 0.034078583121299744, 0.028373030945658684, -0.006847881246358156, -0.02132340520620346, -0.051415521651506424, -0.012872252613306046, 0.007263068575412035, -0.061356429010629654, ...
Compression of Deep Convolutional Neural Networks for Fast and Low Power Mobile Applications
https://arxiv.org/abs/1511.06530
[ "Yong-Deok Kim", "Eunhyeok Park", "Sungjoo Yoo", "Taelim Choi", "Lu Yang", "Dongjun Shin" ]
null
null
Although the latest high-end smartphone has powerful CPU and GPU, running deeper convolutional neural networks (CNNs) for complex tasks such as ImageNet classification on mobile devices is challenging. To deploy deep CNNs on mobile devices, we present a simple and effective scheme to compress the entire CNN, which we...
[]
null
37
1511.06530
iclr_archive
[ -0.017092199996113777, -0.05640258267521858, -0.0005343892844393849, 0.01825873553752899, 0.06460589170455933, 0.058213911950588226, 0.011194038204848766, 0.0017513196216896176, -0.010495205409824848, -0.05987969785928726, 0.007941527292132378, -0.03390437364578247, -0.05275034159421921, -...
Neural Programmer-Interpreters
https://arxiv.org/abs/1511.06279
[ "Scott Reed", "Nando de Freitas" ]
null
null
We propose the neural programmer-interpreter (NPI): a recurrent and compositional neural network that learns to represent and execute programs. NPI has three learnable components: a task-agnostic recurrent core, a persistent key-value program memory, and domain-specific encoders that enable a single NPI to operate in...
[]
null
38
1511.06279
iclr_archive
[ -0.017024751752614975, -0.004174697212874889, -0.04109878093004227, 0.022213827818632126, 0.028643086552619934, 0.057844482362270355, 0.012532509863376617, 0.011070717126131058, -0.04210064560174942, -0.025139102712273598, -0.026938939467072487, 0.022130975499749184, -0.06201518699526787, ...
Session-based recommendations with recurrent neural networks
https://arxiv.org/abs/1511.06939
[ "[code] Balázs Hidasi", "Alexandros Karatzoglou", "Linas Baltrunas", "Domonkos Tikk" ]
null
null
We apply recurrent neural networks (RNN) on a new domain, namely recommender systems. Real-life recommender systems often face the problem of having to base recommendations only on short session-based data (e.g. a small sportsware website) instead of long user histories (as in the case of Netflix). In this situation ...
[]
null
39
1511.06939
iclr_archive
[ 0.015392093919217587, -0.044786155223846436, -0.003039185656234622, 0.04830546677112579, 0.0649830624461174, 0.003934832289814949, 0.029360054060816765, 0.016547946259379387, -0.009495665319263935, -0.04003995284438133, -0.024745607748627663, 0.019254222512245178, -0.05258693918585777, -0....
Continuous control with deep reinforcement learning
https://arxiv.org/abs/1509.02971
[ "Timothy Lillicrap", "Jonathan Hunt", "Alexander Pritzel", "Nicolas Heess", "Tom Erez", "Yuval Tassa", "David Silver", "Daan Wierstra" ]
null
null
We adapt the ideas underlying the success of Deep Q-Learning to the continuous action domain. We present an actor-critic, model-free algorithm based on the deterministic policy gradient that can operate over continuous action spaces. Using the same learning algorithm, network architecture and hyper-parameters, our al...
[]
null
40
1509.02971
iclr_archive
[ -0.02387125790119171, -0.03058590367436409, -0.018422435969114304, 0.05960315838456154, 0.03577234596014023, 0.011013061739504337, 0.007949721068143845, -0.008138350211083889, -0.015752317383885384, -0.031058797612786293, -0.02104528620839119, 0.018074598163366318, -0.0682060495018959, -0....
Recurrent Gaussian Processes
https://arxiv.org/abs/1511.06644
[ "César Lincoln Mattos", "Zhenwen Dai", "Andreas Damianou", "Jeremy Forth", "Guilherme Barreto", "Neil Lawrence" ]
null
null
We define Recurrent Gaussian Processes (RGP) models, a general family of Bayesian nonparametric models with recurrent GP priors which are able to learn dynamical patterns from sequential data. Similar to Recurrent Neural Networks (RNNs), RGPs can have different formulations for their internal states, distinct inferen...
[]
null
41
1511.06644
iclr_archive
[ 0.01981847919523716, -0.00889674387872219, 0.01156180165708065, 0.03848707675933838, 0.03193533048033714, 0.05869961157441139, 0.03966861218214035, 0.0263922531157732, -0.039791449904441833, -0.04927023872733116, -0.024557488039135933, -0.0029903831891715527, -0.06524926424026489, -0.01280...
Modeling Visual Representations:Defining Properties and Deep Approximations
https://arxiv.org/abs/1411.7676
[ "Stefano Soatto", "Alessandro Chiuso" ]
null
null
Visual representations are defined in terms of minimal sufficient statistics of visual data, for a class of tasks, that are also invariant to nuisance variability. Minimal sufficiency guarantees that we can store a representation in lieu of raw data with smallest complexity and no performance loss on the task at hand...
[]
null
42
1411.7676
iclr_archive
[ 0.007112422958016396, 0.0064336881041526794, 0.01878931373357773, 0.03043207712471485, 0.027529040351510048, 0.04159732535481453, 0.0052037774585187435, 0.0055202278308570385, -0.04634687677025795, -0.02619773894548416, -0.05609649792313576, -0.013648636639118195, -0.08383522182703018, 0.0...
Auxiliary Image Regularization for Deep CNNs with Noisy Labels
https://arxiv.org/abs/1511.07069
[ "Samaneh Azadi", "Jiashi Feng", "Stefanie Jegelka", "Trevor Darrell" ]
null
null
Precisely-labeled data sets with sufficient amount of samples are very important for training deep convolutional neural networks (CNNs). However, many of the available real-world data sets contain erroneously labeled samples and those errors substantially hinder the learning of very accurate CNN models. In this work,...
[]
null
43
1511.07069
iclr_archive
[ 0.013181847520172596, -0.03224228695034981, -0.03731102868914604, 0.031354304403066635, 0.014556570909917355, 0.04823723062872887, 0.017454762011766434, -0.027599146589636803, -0.04993332177400589, -0.06257757544517517, -0.019738972187042236, 0.0031764593441039324, -0.050057049840688705, -...
Convergent Learning: Do different neural networks learn the same representations?
https://arxiv.org/abs/1511.07543
[ "Yixuan Li", "Jason Yosinski", "Jeff Clune", "Hod Lipson", "John Hopcroft" ]
null
null
Recent success in training deep neural networks have prompted active investigation into the features learned on their intermediate layers. Such research is difficult because it requires making sense of non-linear computations performed by millions of parameters, but valuable because it increases our ability to unders...
[]
null
44
1511.07543
iclr_archive
[ -0.021159542724490166, -0.028655199334025383, -0.007751526776701212, 0.041248809546232224, 0.039724528789520264, 0.015500952489674091, 0.02938777208328247, 0.01738661155104637, -0.03628232702612877, -0.04008633270859718, 0.0064898827113211155, -0.01860024407505989, -0.06226203963160515, 0....
Policy Distillation
https://arxiv.org/abs/1511.06295
[ "Andrei Rusu", "Sergio Gomez", "Caglar Gulcehre", "Guillaume Desjardins", "James Kirkpatrick", "Razvan Pascanu", "Volodymyr Mnih", "Koray Kavukcuoglu", "Raia Hadsell" ]
null
null
Policies for complex visual tasks have been successfully learned with deep reinforcement learning, using an approach called deep Q-networks (DQN), but relatively large (task-specific) networks and extensive training are needed to achieve good performance. In this work, we present a novel method called policy distilla...
[]
null
45
1511.06295
iclr_archive
[ 0.00410880520939827, -0.02095801942050457, -0.022590599954128265, 0.06973125040531158, 0.03928188234567642, -0.0005116355023346841, 0.01424854900687933, -0.024710718542337418, -0.022505145519971848, -0.0237562395632267, -0.03215312585234642, 0.008876219391822815, -0.08448699861764908, -0.0...
Neural Random-Access Machines
https://arxiv.org/abs/1511.06392
[ "Karol Kurach", "Marcin Andrychowicz", "Ilya Sutskever" ]
null
null
In this paper, we propose and investigate a new neural network architecture called Neural Random Access Machine. It can manipulate and dereference pointers to an external variable-size random-access memory. The model is trained from pure input-output examples using backpropagation. We evaluate the new model on a nu...
[]
null
46
1511.06392
iclr_archive
[ -0.017138520255684853, 0.030431680381298065, -0.03334622457623482, 0.042349398136138916, 0.015736408531665802, 0.04447286203503609, 0.01262948289513588, 0.028165016323328018, -0.05315640568733215, -0.02627493441104889, 0.0009491952951066196, -0.007937871851027012, -0.0509425587952137, -0.0...
Gated Graph Sequence Neural Networks
https://arxiv.org/abs/1511.05493
[ "Yujia Li", "Daniel Tarlow", "Marc Brockschmidt", "Richard Zemel", "CIFAR" ]
null
null
Graph-structured data appears frequently in domains including chemistry, natural language semantics, social networks, and knowledge bases. In this work, we study feature learning techniques for graph-structured inputs. Our starting point is previous work on Graph Neural Networks (Scarselli et al., 2009), which we mod...
[]
null
47
1511.05493
iclr_archive
[ -0.014521097764372826, -0.02548193745315075, -0.006568658631294966, 0.06498942524194717, 0.029353570193052292, 0.013726856559515, 0.02272537164390087, 0.012001962400972843, 0.023944087326526642, -0.027107225731015205, 0.02683348022401333, 0.004192591179162264, -0.0709376409649849, 0.003291...
Metric Learning with Adaptive Density Discrimination
https://arxiv.org/abs/1511.05939
[ "Oren Rippel", "Manohar Paluri", "Piotr Dollar", "Lubomir Bourdev" ]
null
null
Distance metric learning (DML) approaches learn a transformation to a representation space where distance is in correspondence with a predefined notion of similarity. While such models offer a number of compelling benefits, it has been difficult for these to compete with modern classification algorithms in performanc...
[]
null
48
1511.05939
iclr_archive
[ 0.007872364483773708, 0.008405604399740696, 0.0007859900943003595, 0.024716654792428017, 0.035771384835243225, 0.03881458938121796, -0.001629940583370626, -0.009401394985616207, -0.028039980679750443, -0.02705528400838375, -0.026568034663796425, 0.0063105057924985886, -0.08364470303058624, ...
Censoring Representations with an Adversary
https://arxiv.org/abs/1511.05897
[ "Harrison Edwards", "Amos Storkey" ]
null
null
In practice, there are often explicit constraints on what representations or decisions are acceptable in an application of machine learning. For example it may be a legal requirement that a decision must not favour a particular group. Alternatively it can be that that representation of data must not have identifying ...
[]
null
49
1511.05897
iclr_archive
[ -0.009603838436305523, -0.015149622224271297, -0.030882256105542183, 0.06314833462238312, 0.01877482607960701, 0.022507386282086372, 0.018563564866781235, -0.03362486511468887, -0.038236718624830246, -0.03370748460292816, -0.033555082976818085, 0.021966511383652687, -0.06669284403324127, 0...
Order-Embeddings of Images and Language
https://arxiv.org/abs/1511.06361
[ "[code] Ivan Vendrov", "Ryan Kiros", "Sanja Fidler", "Raquel Urtasun" ]
null
null
Hypernymy, textual entailment, and image captioning can be seen as special cases of a single visual-semantic hierarchy over words, sentences, and images. In this paper we advocate for explicitly modeling the partial order structure of this hierarchy. Towards this goal, we introduce a general method for learning order...
[]
null
50
1511.06361
iclr_archive
[ -0.031282585114240646, 0.0006480178562924266, 0.004933524876832962, 0.05134822800755501, 0.018816959112882614, -0.005552250891923904, 0.04181530699133873, -0.00021311166347004473, -0.03359438478946686, -0.0013575105695053935, -0.032314855605363846, 0.035492245107889175, -0.053854990750551224...
Variable Rate Image Compression with Recurrent Neural Networks
https://arxiv.org/abs/1511.06085
[ "George Toderici", "Sean O'Malley", "Damien Vincent", "Sung Jin Hwang", "Michele Covell", "Shumeet Baluja", "Rahul Sukthankar", "David Minnen" ]
null
null
A large fraction of Internet traffic is now driven by requests from mobile devices with relatively small screens and often stringent bandwidth requirements. Due to these factors, it has become the norm for modern graphics-heavy websites to transmit low-resolution, low-bytecount image previews (thumbnails) as part of ...
[]
null
51
1511.06085
iclr_archive
[ 0.0026713982224464417, -0.05219888687133789, -0.014834619127213955, 0.03816535696387291, 0.054219819605350494, 0.04499165713787079, 0.024818146601319313, 0.02090628072619438, -0.033865898847579956, -0.054533377289772034, -0.01854599267244339, -0.035238251090049744, -0.04699608311057091, 0....
Delving Deeper into Convolutional Networks for Learning Video Representations
https://arxiv.org/abs/1511.06432
[ "Nicolas Ballas", "Li Yao", "Pal Chris", "Aaron Courville" ]
null
null
We propose an approach to learn spatio-temporal features in videos from intermediate visual representations we call "percepts" using Gated-Recurrent-Unit Recurrent Networks (GRUs).Our method relies on percepts that are extracted from all level of a deep convolutional network trained on the large ImageNet dataset. Whi...
[]
null
52
1511.06432
iclr_archive
[ 0.028864730149507523, -0.013468162156641483, 0.01659388281404972, 0.056404031813144684, 0.016760988160967827, 0.03731894865632057, 0.026073656976222992, 0.03847571089863777, -0.007863779552280903, -0.0378788523375988, -0.0011384680401533842, -0.029276998713612556, -0.07292444258928299, 0.0...
8-Bit Approximations for Parallelism in Deep Learning
https://arxiv.org/abs/1511.04561
[ "Tim Dettmers" ]
null
null
The creation of practical deep learning data-products often requires parallelization across processors and computers to make deep learning feasible on large data sets, but bottlenecks in communication bandwidth make it difficult to attain good speedups through parallelism. Here we develop and test 8-bit approximation...
[]
null
53
1511.04561
iclr_archive
[ -0.020023196935653687, -0.0377233624458313, -0.013996272347867489, 0.040932547301054, 0.03883964940905571, 0.04752257093787193, -0.00192011590115726, 0.016143186017870903, -0.02901439368724823, -0.029823610559105873, 0.011518844403326511, -0.02210915833711624, -0.06368737667798996, 0.01926...
Data-dependent initializations of Convolutional Neural Networks
https://arxiv.org/abs/1511.06856
[ "[code] Philipp Kraehenbuehl", "Carl Doersch", "Jeff Donahue", "Trevor Darrell" ]
null
null
Convolutional Neural Networks spread through computer vision like a wildfire, impacting almost all visual tasks imaginable. Despite this, few researchers dare to train their models from scratch. Most work builds on one of a handful of ImageNet pre-trained models, and fine-tunes or adapts these for specific tasks. Thi...
[]
null
54
1511.06856
iclr_archive
[ 0.02043193392455578, -0.024489378556609154, 0.01096490491181612, 0.04279365763068199, 0.04336334392428398, 0.04506855830550194, -0.023682767525315285, -0.001525853993371129, -0.010859835892915726, -0.03743901476264, -0.030677085742354393, 0.004591816570609808, -0.06808749586343765, -0.0057...
Order Matters: Sequence to sequence for sets
https://arxiv.org/abs/1511.06391
[ "Oriol Vinyals", "Samy Bengio", "Manjunath Kudlur" ]
null
null
Sequences have become first class citizens in supervised learning thanks to the resurgence of recurrent neural networks. Many complex tasks that require mapping from or to a sequence of observations can now be formulated with the sequence-to-sequence (seq2seq) framework which employs the chain rule to efficiently rep...
[]
null
55
1511.06391
iclr_archive
[ -0.020914660766720772, -0.02423199452459812, -0.036747127771377563, 0.028533613309264183, 0.024688150733709335, 0.035401925444602966, 0.015736231580376625, 0.01379283145070076, -0.023404356092214584, -0.024425551295280457, -0.01332465186715126, 0.04875844717025757, -0.07487412542104721, -0...
High-Dimensional Continuous Control Using Generalized Advantage Estimation
https://arxiv.org/abs/1506.02438
[ "John Schulman", "Philipp Moritz", "Sergey Levine", "Michael Jordan", "Pieter Abbeel" ]
null
null
Policy gradient methods are an appealing approach in reinforcement learning because they directly optimize the cumulative reward and can straightforwardly be used with nonlinear function approximators such as neural networks. The two main challenges are the large number of samples typically required, and the difficul...
[]
null
56
1506.02438
iclr_archive
[ -0.01786387339234352, -0.027228454127907753, -0.007791643496602774, 0.05624489113688469, 0.03271942585706711, 0.03730729594826698, 0.028275124728679657, -0.02545223943889141, -0.0176171213388443, -0.046701595187187195, 0.006413978058844805, 0.0047613768838346004, -0.09155106544494629, -0.0...
BlackOut: Speeding up Recurrent Neural Network Language Models With Very Large Vocabularies
https://arxiv.org/abs/1511.06909
[ "[code] Shihao Ji", "Swaminathan Vishwanathan", "Nadathur Satish", "Michael Anderson", "Pradeep Dubey" ]
null
null
We propose BlackOut, an approximation algorithm to efficiently train massive recurrent neural network language models (RNNLMs) with million word vocabularies. BlackOut is motivated by using a discriminative loss, and we describe a new sampling strategy which significantly reduces computation while improving stability...
[]
null
57
1511.06909
iclr_archive
[ -0.013291231356561184, -0.033780086785554886, -0.01592722348868847, 0.04915761202573776, 0.0250905379652977, 0.042579248547554016, 0.03060770221054554, 0.01826154999434948, -0.04768771305680275, -0.0014500682009384036, -0.0333724170923233, 0.021444004029035568, -0.06277702003717422, -0.004...
Deep Multi Scale Video Prediction Beyond Mean Square Error
https://arxiv.org/abs/1511.05440
[ "Michael Mathieu", "camille couprie", "Yann Lecun" ]
null
null
Learning to predict future images from a video sequence involves the construction of an internal representation that models the image evolution accurately, and therefore, to some degree, its content and dynamics. This is why pixel-space video prediction may be viewed as a promising avenue for unsupervised feature lea...
[]
null
58
1511.05440
iclr_archive
[ 0.011604350060224533, -0.0026383530348539352, 0.02907823771238327, 0.0377371683716774, 0.04213060811161995, 0.04706821218132973, 0.01687707006931305, 0.015580102801322937, -0.04005572199821472, -0.05082210898399353, 0.0005966852186247706, -0.018105149269104004, -0.05554087460041046, 0.0144...
Grid Long Short-Term Memory
https://arxiv.org/abs/1507.01526
[ "Nal Kalchbrenner", "Alex Graves", "Ivo Danihelka" ]
null
null
This paper introduces Grid Long Short-Term Memory, a network of LSTM cells arranged in a multidimensional grid that can be applied to vectors, sequences or higher dimensional data such as images. The network differs from existing deep LSTM architectures in that the cells are connected between network layers as well a...
[]
null
59
1507.01526
iclr_archive
[ -0.030579660087823868, -0.02144157513976097, -0.016871431842446327, 0.04560793563723564, 0.032523151487112045, 0.03776130452752113, 0.013930295594036579, 0.029780155047774315, -0.04844702407717705, -0.0267090555280447, 0.0008076227968558669, -0.04204433783888817, -0.055549245327711105, 0.0...
Net2Net: Accelerating Learning via Knowledge Transfer
https://arxiv.org/abs/1511.05641
[ "Tianqi Chen", "Ian Goodfellow", "Jon Shlens" ]
null
null
We introduce techniques for rapidly transferring the information stored in one neural net into another neural net. The main purpose is to accelerate the training of a significantly larger neural net. During real-world workflows, one often trains very many different neural networks during the experimentation and desig...
[]
null
60
1511.05641
iclr_archive
[ 0.01090665441006422, -0.05981398746371269, -0.014359494671225548, 0.029471395537257195, 0.044557731598615646, 0.03160594776272774, -0.008958471938967705, -0.02304975315928459, 0.0024839136749505997, -0.03583718463778496, -0.00585662666708231, -0.003330869833007455, -0.039339084178209305, 0...
Predicting distributions with Linearizing Belief Networks
https://arxiv.org/abs/1511.05622
[ "Yann Dauphin", "David Grangier" ]
null
null
Conditional belief networks introduce stochastic binary variables in neural networks. Contrary to a classical neural network, a belief network can predict more than the expected value of the output $Y$ given the input $X$. It can predict a distribution of outputs $Y$ which is useful when an input can admit multiple o...
[]
null
61
1511.05622
iclr_archive
[ -0.00808250717818737, 0.009204873815178871, -0.017425181344151497, 0.03592204675078392, 0.04820206016302109, 0.07698952406644821, 0.02830459736287594, -0.0025166175328195095, -0.019812993705272675, -0.05803997442126274, 0.0028776153922080994, 0.006275642663240433, -0.06372066587209702, -0....
Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
https://arxiv.org/abs/1511.07289
[ "Djork-Arné Clevert", "Thomas Unterthiner", "Sepp Hochreiter" ]
null
null
We introduce the "exponential linear unit" (ELU) which speeds up learning in deep neural networks and leads to higher classification accuracies. Like rectified linear units (ReLUs), leaky ReLUs (LReLUs) and parametrized ReLUs (PReLUs), ELUs alleviate the vanishing gradient problem via the identity for positive values...
[]
null
62
1511.07289
iclr_archive
[ 0.004401254002004862, -0.013173508457839489, -0.0017590399365872145, 0.039751313626766205, 0.03788149356842041, 0.04730265960097313, 0.02028834819793701, 0.010557360015809536, -0.018855050206184387, -0.03942398354411125, -0.00701186852529645, -0.01924661174416542, -0.07195574045181274, -0....
Actor-Mimic: Deep Multitask and Transfer Reinforcement Learning
https://arxiv.org/abs/1511.06342
[ "Emilio Parisotto", "Jimmy Ba", "Ruslan Salakhutdinov" ]
null
null
The ability to act in multiple environments and transfer previous knowledge to new situations can be considered a critical aspect of any intelligent agent. Towards this goal, we define a novel method of multitask and transfer learning that enables an autonomous agent to learn how to behave in multiple tasks simultane...
[]
null
63
1511.06342
iclr_archive
[ -0.0147046884521842, -0.03715483471751213, -0.02190006896853447, 0.025734150782227516, 0.06023481860756874, 0.03697696700692177, 0.011564599350094795, 0.0013762896414846182, -0.02036842331290245, -0.043629713356494904, -0.006633301265537739, 0.01730639487504959, -0.07176283746957779, -0.03...
Segmental Recurrent Neural Networks
https://arxiv.org/abs/1511.06018
[ "Lingpeng Kong", "Chris Dyer", "Noah Smith" ]
null
null
We introduce segmental recurrent neural networks (SRNNs) which define, given an input sequence, a joint probability distribution over segmentations of the input and labelings of the segments. Representations of the input segments (i.e., contiguous subsequences of the input) are computed by encoding their constituent ...
[]
null
64
1511.06018
iclr_archive
[ -0.024789931252598763, -0.006668256130069494, -0.03235792741179466, 0.02915591560304165, 0.040135689079761505, 0.03664352372288704, 0.013561105355620384, 0.038549523800611496, -0.029683643952012062, -0.02330872416496277, -0.0205007903277874, -0.013050222769379616, -0.050571147352457047, -0...
Deep Linear Discriminant Analysis
https://arxiv.org/abs/1511.04707
[ "[code] Matthias Dorfer", "Rainer Kelz", "Gerhard Widmer" ]
null
null
We introduce Deep Linear Discriminant Analysis (DeepLDA) which learns linearly separable latent representations in an end-to-end fashion. Classic LDA extracts features which preserve class separability and is used for dimensionality reduction for many classification problems. The central idea of this paper is to put ...
[]
null
65
1511.04707
iclr_archive
[ 0.012208878062665462, -0.0020135832019150257, -0.008541347458958626, 0.055701784789562225, 0.04875057563185692, 0.027075009420514107, 0.01294772233814001, -0.016950460150837898, 0.020255958661437035, -0.039830513298511505, -0.02696889452636242, -0.003314871108159423, -0.07296614348888397, ...
Large-Scale Approximate Kernel Canonical Correlation Analysis
https://arxiv.org/abs/1511.04773
[ "Weiran Wang", "Karen Livescu" ]
null
null
Kernel canonical correlation analysis (KCCA) is a nonlinear multi-view representation learning technique with broad applicability in statistics and machine learning. Although there is a closed-form solution for the KCCA objective, it involves solving an $N\times N$ eigenvalue system where $N$ is the training set size...
[]
null
66
1511.04773
iclr_archive
[ -0.02605637162923813, -0.016015054658055305, 0.0021745592821389437, 0.03412434086203575, 0.035173289477825165, 0.053559672087430954, 0.029321761801838875, 0.010754087939858437, -0.009354420937597752, -0.019929220899939537, -0.013593717478215694, -0.008360902778804302, -0.07522817701101303, ...
Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
https://arxiv.org/abs/1511.06434
[ "Alec Radford", "Luke Metz", "Soumith Chintala" ]
null
null
In recent years, supervised learning with convolutional networks (CNNs) has seen huge adoption in computer vision applications. Comparatively, unsupervised learning with CNNs has received less attention. In this work we hope to help bridge the gap between the success of CNNs for supervised learning and unsupervised l...
[]
null
67
1511.06434
iclr_archive
[ 0.018260760232806206, -0.030348965898156166, -0.016494981944561005, 0.04505213350057602, 0.020305464044213295, 0.005650591570883989, -0.005951328668743372, 0.00762441661208868, -0.011521643958985806, -0.04452770575881004, -0.018023451790213585, 0.004973947536200285, -0.05319136753678322, 0...
Learning Representations from EEG with Deep Recurrent-Convolutional Neural Networks
https://arxiv.org/abs/1511.06448
[ "[code] Pouya Bashivan", "Irina Rish", "Mohammed Yeasin", "Noel Codella" ]
null
null
One of the challenges in modeling cognitive events from electroencephalogram (EEG) data is finding representations that are invariant to inter- and intra-subject differences, as well as to inherent noise associated with such data. Herein, we propose a novel approach for learning such representations from multi-channe...
[]
null
68
1511.06448
iclr_archive
[ 0.004930460825562477, -0.0009531069081276655, 0.0017528344178572297, 0.005017525982111692, 0.036038853228092194, 0.043983329087495804, 0.042638782411813736, 0.008933475241065025, 0.0005780212813988328, -0.05020511895418167, -0.018902719020843506, -0.00010229517647530884, -0.07720334082841873...
Digging Deep into the layers of CNNs: In Search of How CNNs Achieve View Invariance
https://arxiv.org/abs/1508.01983
[ "Amr Bakry", "Mohamed Elhoseiny", "Tarek El-Gaaly", "Ahmed Elgammal" ]
null
null
This paper is focused on studying the view-manifold structure in the feature spaces implied by the different layers of Convolutional Neural Networks (CNN). There are several questions that this paper aims to answer: Does the learned CNN representation achieve viewpoint invariance? How does it achieve viewpoint invari...
[]
null
69
1508.01983
iclr_archive
[ 0.0021490012295544147, -0.007368990685790777, 0.005699708592146635, 0.009653358720242977, 0.03527321293950081, 0.02621474117040634, 0.021818704903125763, 0.006853281520307064, -0.013357921503484249, -0.04678117856383324, -0.028908858075737953, -0.016776902601122856, -0.06432332843542099, 0...
An Exploration of Softmax Alternatives Belonging to the Spherical Loss Family
https://arxiv.org/abs/1511.05042
[ "Alexandre De Brébisson", "Pascal Vincent" ]
null
null
In a multi-class classification problem, it is standard to model the output of a neural network as a categorical distribution conditioned on the inputs. The output must therefore be positive and sum to one, which is traditionally enforced by a softmax. This probabilistic mapping allows to use the maximum likelihood p...
[]
null
70
1511.05042
iclr_archive
[ -0.03689706698060036, 0.0020206086337566376, 0.021915823221206665, -0.005648867227137089, 0.0313742570579052, 0.03850354626774788, 0.014931091107428074, -0.017997024580836296, -0.025358961895108223, -0.04305574670433998, -0.01600942388176918, 0.02290390431880951, -0.07499616593122482, -0.0...
Data-Dependent Path Normalization in Neural Networks
https://arxiv.org/abs/1511.06747
[ "Behnam Neyshabur", "Ryota Tomioka", "Ruslan Salakhutdinov", "Nathan Srebro" ]
null
null
We propose a unified framework for neural net normalization, regularization and optimization, which includes Path-SGD and Batch-Normalization and interpolates between them across two different dimensions. Through this framework we investigate issue of invariance of the optimization, data dependence and the connection...
[]
null
71
1511.06747
iclr_archive
[ -0.011797702871263027, -0.022287961095571518, -0.006633341312408447, 0.03424140065908432, 0.0628749206662178, 0.06110488995909691, 0.018004318699240685, -0.006077548023313284, -0.04635041952133179, -0.06317518651485443, 0.006810799706727266, -0.008779145777225494, -0.04385809600353241, -0....
Reasoning in Vector Space: An Exploratory Study of Question Answering
https://arxiv.org/abs/1511.06426
[ "Moontae Lee", "Xiaodong He", "Wen-tau Yih", "Jianfeng Gao", "Li Deng", "Paul Smolensky" ]
null
null
Question answering tasks have shown remarkable progress with distributed vector representation. In this paper, we investigate the recently proposed Facebook bAbI tasks which consist of twenty different categories of questions that require complex reasoning. Because the previous work on bAbI are all end-to-end models,...
[]
null
72
1511.06426
iclr_archive
[ 0.00012240871728863567, -0.012523835524916649, 0.0007609293679706752, 0.04465886577963829, 0.04174615070223808, -0.0015717096393927932, 0.018007894977927208, 0.002341588493436575, -0.0002708809915930033, -0.005552325397729874, -0.03063611499965191, 0.01594206877052784, -0.07490593940019608, ...
Neural GPUs Learn Algorithms
https://arxiv.org/abs/1511.08228
[ "[code] [video] Lukasz Kaiser", "Ilya Sutskever" ]
null
null
Learning an algorithm from examples is a fundamental problem that has been widely studied. Recently it has been addressed using neural networks, in particular by Neural Turing Machines (NTMs). These are fully differentiable computers that use backpropagation to learn their own programming. Despite their appeal NTMs h...
[]
null
73
1511.08228
iclr_archive
[ -0.004478097427636385, -0.009232092648744583, -0.004157849121838808, 0.04404144734144211, 0.018042372539639473, 0.05305517837405205, 0.009912019595503807, 0.054053932428359985, -0.03156784549355507, -0.024548858404159546, -0.0011995796812698245, 0.006937025114893913, -0.05992033705115318, ...
ACDC: A Structured Efficient Linear Layer
https://arxiv.org/abs/1511.05946
[ "Marcin Moczulski", "Misha Denil", "Jeremy Appleyard", "Nando de Freitas" ]
null
null
The linear layer is one of the most pervasive modules in deep learning representations. However, it requires $O(N^2)$ parameters and $O(N^2)$ operations. These costs can be prohibitive in mobile applications or prevent scaling in many domains. Here, we introduce a deep, differentiable, fully-connected neural network ...
[]
null
74
1511.05946
iclr_archive
[ -0.010542670264840126, 0.00514606200158596, 0.0017575762467458844, 0.03198631852865219, 0.0448601059615612, 0.02108287811279297, -0.011387995444238186, -0.001441202126443386, -0.02458622120320797, -0.03741078078746796, -0.01722533255815506, -0.02366439439356327, -0.054361745715141296, 0.00...
Density Modeling of Images using a Generalized Normalization Transformation
https://arxiv.org/abs/1511.06281
[ "Johannes Ballé", "Valero Laparra", "Eero Simoncelli" ]
null
null
We introduce a parametric nonlinear transformation that is well-suited for Gaussianizing data from natural images. The data are linearly transformed, and each component is then normalized by a pooled activity measure, computed by exponentiating a weighted sum of rectified and exponentiated components and a constant. ...
[]
null
75
1511.06281
iclr_archive
[ 0.0010950511787086725, -0.014534329995512962, 0.015041871927678585, 0.028635939583182335, 0.04197186231613159, 0.05551944300532341, 0.017725056037306786, -0.0010067189577966928, -0.04252823814749718, -0.04883556813001633, -0.025475382804870605, -0.0170915350317955, -0.049862202256917953, -...
Adversarial Manipulation of Deep Representations
https://arxiv.org/abs/1511.05122
[ "[code] Sara Sabour", "Yanshuai Cao", "Fartash Faghri", "David Fleet" ]
null
null
We show that the representation of an image in a deep neural network (DNN) can be manipulated to mimic those of other natural images, with only minor, imperceptible perturbations to the original image. Previous methods for generating adversarial images focused on image perturbations designed to produce erroneous clas...
[]
null
76
1511.05122
iclr_archive
[ 0.0061780232936143875, -0.009691739454865456, -0.03198814392089844, 0.05924477055668831, 0.030239678919315338, 0.021219823509454727, 0.021773312240839005, -0.016119612380862236, -0.03390364721417427, -0.042341768741607666, -0.048674650490283966, -0.016330348327755928, -0.05250536650419235, ...
Geodesics of learned representations
https://arxiv.org/abs/1511.06394
[ "Olivier Hénaff", "Eero Simoncelli" ]
null
null
We develop a new method for visualizing and refining the invariances of learned representations. Specifically, we test for a general form of invariance, linearization, in which the action of a transformation is confined to a low-dimensional subspace. Given two reference images (typically, differing by some transforma...
[]
null
77
1511.06394
iclr_archive
[ -0.027192987501621246, 0.012899807654321194, 0.009373641572892666, 0.02263614349067211, 0.031909335404634476, 0.022706352174282074, 0.022894687950611115, 0.035677265375852585, -0.05773557722568512, -0.03911840543150902, -0.01794447936117649, -0.04486401006579399, -0.05990774556994438, 0.01...
Sequence Level Training with Recurrent Neural Networks
https://arxiv.org/abs/1511.06732
[ "Marc'Aurelio Ranzato", "Sumit Chopra", "Michael Auli", "Wojciech Zaremba" ]
null
null
Many natural language processing applications use language models to generate text. These models are typically trained to predict the next word in a sequence, given the previous words and some context such as an image. However, at test time the model is expected to generate the entire sequence from scratch. This disc...
[]
null
78
1511.06732
iclr_archive
[ -0.010025429539382458, -0.019796235486865044, -0.02642587013542652, 0.05562719702720642, 0.05172228813171387, 0.01993553712964058, 0.03747877851128578, 0.03473779186606407, -0.029828524217009544, -0.00425568874925375, -0.015749596059322357, 0.03939025104045868, -0.04685862362384796, -0.006...
Super-resolution with deep convolutional sufficient statistics
https://arxiv.org/abs/1511.05666
[ "Joan Bruna", "Pablo Sprechmann", "Yann Lecun" ]
null
null
Inverse problems in image and audio, and super-resolution in particular, can be seen as high-dimensional structured prediction problems, where the goal is to characterize the conditional distribution of a high-resolution output given its low-resolution corrupted observation. When the scaling ratio is small, point est...
[]
null
79
1511.05666
iclr_archive
[ -0.012306218035519123, -0.01619678921997547, -0.0041568991728127, 0.023803161457180977, 0.05668705329298973, 0.050463587045669556, 0.019568882882595062, -0.024479839950799942, -0.03056362271308899, -0.07003657519817352, -0.00461170356720686, 0.018833624199032784, -0.05384916812181473, 0.02...
Variational Gaussian Process
https://arxiv.org/abs/1511.06499
[ "Dustin Tran", "Rajesh Ranganath", "David Blei" ]
null
null
Variational inference is a powerful tool for approximate inference, and it has been recently applied for representation learning with deep generative models. We develop the variational Gaussian process (VGP), a Bayesian nonparametric variational family, which adapts its shape to match complex posterior distributions....
[]
null
80
1511.06499
iclr_archive
[ 0.016261683776974678, 0.016947491094470024, 0.008608393371105194, 0.0468999519944191, 0.02068485878407955, 0.05858607962727547, 0.023597879335284233, 0.01110734511166811, -0.01577882468700409, -0.048966918140649796, -0.03332630917429924, 0.005196025129407644, -0.06839782744646072, 0.024543...