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
Regularized Learning for Domain Adaptation under Label Shifts
https://openreview.net/forum?id=rJl0r3R9KX
[ "Kamyar Azizzadenesheli", "Anqi Liu", "Fanny Yang", "Animashree Anandkumar" ]
Poster
null
We propose Regularized Learning under Label shifts (RLLS), a principled and a practical domain-adaptation algorithm to correct for shifts in the label distribution between a source and a target domain. We first estimate importance weights using labeled source data and unlabeled target data, and then train a classifier ...
[ "Deep Learning", "Domain Adaptation", "Label Shift", "Importance Weights", "Generalization" ]
A practical and provably guaranteed approach for training efficiently classifiers in the presence of label shifts between Source and Target data sets
1,592
1903.09734
title_snapshot
[ -0.008355793543159962, -0.03551222011446953, 0.000540528038982302, 0.03643732890486717, 0.05993608012795448, 0.029751554131507874, 0.012914181686937809, -0.02493065968155861, -0.036190491169691086, -0.006249269004911184, -0.03056429512798786, 0.05140169709920883, -0.07813004404306412, -0.0...
Towards Robust, Locally Linear Deep Networks
https://openreview.net/forum?id=SylCrnCcFX
[ "Guang-He Lee", "David Alvarez-Melis", "Tommi S. Jaakkola" ]
Poster
null
Deep networks realize complex mappings that are often understood by their locally linear behavior at or around points of interest. For example, we use the derivative of the mapping with respect to its inputs for sensitivity analysis, or to explain (obtain coordinate relevance for) a prediction. One key challenge is tha...
[ "robust derivatives", "transparency", "interpretability" ]
A scalable algorithm to establish robust derivatives of deep networks w.r.t. the inputs.
1,591
1907.03207
title_snapshot
[ -0.010214514099061489, -0.0021388090681284666, 0.008418746292591095, 0.03669706732034683, 0.04837864637374878, 0.059539906680583954, 0.019164998084306717, -0.02262430265545845, -0.014953463338315487, -0.043174054473638535, 0.017011357471346855, -0.02485562302172184, -0.054193343967199326, ...
The Limitations of Adversarial Training and the Blind-Spot Attack
https://openreview.net/forum?id=HylTBhA5tQ
[ "Huan Zhang*", "Hongge Chen*", "Zhao Song", "Duane Boning", "Inderjit S. Dhillon", "Cho-Jui Hsieh" ]
Poster
null
The adversarial training procedure proposed by Madry et al. (2018) is one of the most effective methods to defend against adversarial examples in deep neural net- works (DNNs). In our paper, we shed some lights on the practicality and the hardness of adversarial training by showing that the effectiveness (robustness on...
[ "Adversarial Examples", "Adversarial Training", "Blind-Spot Attack" ]
We show that even the strongest adversarial training methods cannot defend against adversarial examples crafted on slightly scaled and shifted test images.
1,584
1901.04684
title_snapshot
[ -0.010875895619392395, -0.036266982555389404, -0.012159335426986217, 0.05345934256911278, 0.016757803037762642, 0.0027403458952903748, 0.058111000806093216, -0.003131832228973508, -0.025561748072504997, -0.041391726583242416, -0.015448399819433689, -0.0009182202047668397, -0.0699156671762466...
Learning to Learn without Forgetting by Maximizing Transfer and Minimizing Interference
https://openreview.net/forum?id=B1gTShAct7
[ "Matthew Riemer", "Ignacio Cases", "Robert Ajemian", "Miao Liu", "Irina Rish", "Yuhai Tu", "and Gerald Tesauro" ]
Poster
null
Lack of performance when it comes to continual learning over non-stationary distributions of data remains a major challenge in scaling neural network learning to more human realistic settings. In this work we propose a new conceptualization of the continual learning problem in terms of a temporally symmetric trade-off ...
[]
null
1,583
1810.11910
title_snapshot
[ -0.024284720420837402, -0.01348612830042839, -0.001499209669418633, 0.014705230481922626, 0.03191881999373436, 0.021499771624803543, 0.018169960007071495, 0.009719732217490673, -0.04014584422111511, -0.027173124253749847, 0.006523634307086468, 0.024294497445225716, -0.049145717173814774, 0...
Global-to-local Memory Pointer Networks for Task-Oriented Dialogue
https://openreview.net/forum?id=ryxnHhRqFm
[ "Chien-Sheng Wu", "Richard Socher", "Caiming Xiong" ]
Poster
null
End-to-end task-oriented dialogue is challenging since knowledge bases are usually large, dynamic and hard to incorporate into a learning framework. We propose the global-to-local memory pointer (GLMP) networks to address this issue. In our model, a global memory encoder and a local memory decoder are proposed to share...
[ "pointer networks", "memory networks", "task-oriented dialogue systems", "natural language processing" ]
GLMP: Global memory encoder (context RNN, global pointer) and local memory decoder (sketch RNN, local pointer) that share external knowledge (MemNN) are proposed to strengthen response generation in task-oriented dialogue.
1,581
1901.04713
title_snapshot
[ -0.02044672518968582, -0.0027987018693238497, -0.0025801556184887886, 0.057665012776851654, 0.02694346196949482, 0.021194027736783028, 0.014421127736568451, 0.008368057198822498, -0.00619416031986475, -0.01939457096159458, -0.028734708204865456, 0.025966906920075417, -0.07483135908842087, ...
Rethinking the Value of Network Pruning
https://openreview.net/forum?id=rJlnB3C5Ym
[ "Zhuang Liu", "Mingjie Sun", "Tinghui Zhou", "Gao Huang", "Trevor Darrell" ]
Poster
null
Network pruning is widely used for reducing the heavy inference cost of deep models in low-resource settings. A typical pruning algorithm is a three-stage pipeline, i.e., training (a large model), pruning and fine-tuning. During pruning, according to a certain criterion, redundant weights are pruned and important weigh...
[ "network pruning", "network compression", "architecture search", "train from scratch" ]
In structured network pruning, fine-tuning a pruned model only gives comparable performance with training it from scratch.
1,580
1810.05270
title_snapshot
[ -0.008405479602515697, -0.04458184912800789, -0.014661682769656181, 0.04762870818376541, 0.038951609283685684, 0.04565274342894554, 0.01822986826300621, 0.002051049144938588, -0.024741046130657196, -0.06285911798477173, 0.008586428128182888, 0.004291679244488478, -0.05541527271270752, -0.0...
Neural TTS Stylization with Adversarial and Collaborative Games
https://openreview.net/forum?id=ByzcS3AcYX
[ "Shuang Ma", "Daniel Mcduff", "Yale Song" ]
Poster
null
The modeling of style when synthesizing natural human speech from text has been the focus of significant attention. Some state-of-the-art approaches train an encoder-decoder network on paired text and audio samples (x_txt, x_aud) by encouraging its output to reconstruct x_aud. The synthesized audio waveform is expected...
[ "Text-To-Speech synthesis", "GANs" ]
a generative adversarial network for style modeling in a text-to-speech system
1,570
null
null
[ -0.0017406573751941323, -0.0318998247385025, -0.03508485481142998, 0.06421258300542831, 0.021945279091596603, 0.0426006019115448, 0.025180688127875328, 0.018695399165153503, -0.014139171689748764, -0.06229449808597565, -0.056928981095552444, 0.028436193242669106, -0.0520663782954216, -0.00...
On the Universal Approximability and Complexity Bounds of Quantized ReLU Neural Networks
https://openreview.net/forum?id=SJe9rh0cFX
[ "Yukun Ding", "Jinglan Liu", "Jinjun Xiong", "Yiyu Shi" ]
Poster
null
Compression is a key step to deploy large neural networks on resource-constrained platforms. As a popular compression technique, quantization constrains the number of distinct weight values and thus reducing the number of bits required to represent and store each weight. In this paper, we study the representation power...
[ "Quantized Neural Networks", "Universial Approximability", "Complexity Bounds", "Optimal Bit-width" ]
This paper proves the universal approximability of quantized ReLU neural networks and puts forward the complexity bound given arbitrary error.
1,567
1802.03646
title_snapshot
[ -0.044808559119701385, -0.01892954483628273, -0.011225943453609943, 0.024501020088791847, 0.04610015079379082, 0.052417658269405365, 0.010766507126390934, -0.011247378773987293, -0.029387271031737328, -0.013200259767472744, -0.007714814972132444, -0.00613827258348465, -0.07839066535234451, ...
Poincare Glove: Hyperbolic Word Embeddings
https://openreview.net/forum?id=Ske5r3AqK7
[ "Alexandru Tifrea*", "Gary Becigneul*", "Octavian-Eugen Ganea*" ]
Poster
null
Words are not created equal. In fact, they form an aristocratic graph with a latent hierarchical structure that the next generation of unsupervised learned word embeddings should reveal. In this paper, justified by the notion of delta-hyperbolicity or tree-likeliness of a space, we propose to embed words in a Cartesian...
[ "word embeddings", "hyperbolic spaces", "poincare ball", "hypernymy", "analogy", "similarity", "gaussian embeddings" ]
We embed words in the hyperbolic space and make the connection with the Gaussian word embeddings.
1,565
1810.06546
title_snapshot
[ -0.013699128292500973, -0.019180674105882645, 0.010924606584012508, 0.04438673332333565, 0.034429483115673065, 0.04218379780650139, 0.03361944481730461, -0.00006726705760229379, -0.004542323760688305, -0.05240834131836891, 0.002227034419775009, 0.007460111752152443, -0.07335811853408813, 0...
Eidetic 3D LSTM: A Model for Video Prediction and Beyond
https://openreview.net/forum?id=B1lKS2AqtX
[ "Yunbo Wang", "Lu Jiang", "Ming-Hsuan Yang", "Li-Jia Li", "Mingsheng Long", "Li Fei-Fei" ]
Poster
null
Spatiotemporal predictive learning, though long considered to be a promising self-supervised feature learning method, seldom shows its effectiveness beyond future video prediction. The reason is that it is difficult to learn good representations for both short-term frame dependency and long-term high-level relations. W...
[]
null
1,564
null
null
[ 0.015623487532138824, -0.007794868666678667, 0.0034593725576996803, 0.027533922344446182, 0.04429279640316963, 0.02670375630259514, 0.02441786229610443, 0.01813993975520134, -0.02736961841583252, -0.026360347867012024, 0.011567243374884129, -0.021583551540970802, -0.042684201151132584, 0.0...
Towards GAN Benchmarks Which Require Generalization
https://openreview.net/forum?id=HkxKH2AcFm
[ "Ishaan Gulrajani", "Colin Raffel", "Luke Metz" ]
Poster
null
For many evaluation metrics commonly used as benchmarks for unconditional image generation, trivially memorizing the training set attains a better score than models which are considered state-of-the-art; we consider this problematic. We clarify a necessary condition for an evaluation metric not to behave this way: esti...
[ "evaluation", "generative adversarial networks", "adversarial divergences" ]
We argue that GAN benchmarks must require a large sample from the model to penalize memorization and investigate whether neural network divergences have this property.
1,563
2001.03653
title_snapshot
[ -0.007221007253974676, -0.018826261162757874, -0.006595574785023928, 0.03384552523493767, 0.007612837012857199, 0.004850645549595356, -0.0006024720496498048, 0.014914372004568577, -0.018774021416902542, -0.04450530186295509, 0.001403281930834055, 0.025264669209718704, -0.04905674234032631, ...
There Are Many Consistent Explanations of Unlabeled Data: Why You Should Average
https://openreview.net/forum?id=rkgKBhA5Y7
[ "Ben Athiwaratkun", "Marc Finzi", "Pavel Izmailov", "Andrew Gordon Wilson" ]
Poster
null
Presently the most successful approaches to semi-supervised learning are based on consistency regularization, whereby a model is trained to be robust to small perturbations of its inputs and parameters. To understand consistency regularization, we conceptually explore how loss geometry interacts with training procedure...
[ "semi-supervised learning", "computer vision", "classification", "consistency regularization", "flatness", "weight averaging", "stochastic weight averaging" ]
Consistency-based models for semi-supervised learning do not converge to a single point but continue to explore a diverse set of plausible solutions on the perimeter of a flat region. Weight averaging helps improve generalization performance.
1,559
1806.05594
title_snapshot
[ 0.02044549025595188, -0.0664139911532402, -0.02996409684419632, 0.0648491159081459, 0.0334555022418499, -0.005765019915997982, 0.037435755133628845, -0.004561367444694042, -0.007420670241117477, -0.04916061460971832, -0.004164114128798246, 0.002448802348226309, -0.08515458554029465, 0.0107...
Synthetic Datasets for Neural Program Synthesis
https://openreview.net/forum?id=ryeOSnAqYm
[ "Richard Shin", "Neel Kant", "Kavi Gupta", "Chris Bender", "Brandon Trabucco", "Rishabh Singh", "Dawn Song" ]
Poster
null
The goal of program synthesis is to automatically generate programs in a particular language from corresponding specifications, e.g. input-output behavior. Many current approaches achieve impressive results after training on randomly generated I/O examples in limited domain-specific languages (DSLs), as with string tra...
[]
null
1,558
1912.12345
title_snapshot
[ -0.007237214595079422, -0.008219133131206036, -0.03515461087226868, 0.05525529012084007, 0.042341526597738266, 0.05241930112242699, 0.030882932245731354, 0.010038265027105808, -0.0062273466028273106, -0.016333868727087975, -0.014252664521336555, 0.025622116401791573, -0.06809163093566895, ...
Stochastic Prediction of Multi-Agent Interactions from Partial Observations
https://openreview.net/forum?id=r1xdH3CcKX
[ "Chen Sun", "Per Karlsson", "Jiajun Wu", "Joshua B Tenenbaum", "Kevin Murphy" ]
Poster
null
We present a method which learns to integrate temporal information, from a learned dynamics model, with ambiguous visual information, from a learned vision model, in the context of interacting agents. Our method is based on a graph-structured variational recurrent neural network, which is trained end-to-end to infer th...
[ "Dynamics modeling", "partial observations", "multi-agent interactions", "predictive models" ]
We present a method which learns to integrate temporal information and ambiguous visual information in the context of interacting agents.
1,557
1902.09641
title_snapshot
[ 0.005340706557035446, 0.0047561656683683395, -0.0025902693159878254, 0.029072130098938942, 0.029684454202651978, 0.01913551241159439, 0.044140055775642395, 0.030744075775146484, -0.027774089947342873, -0.04424622654914856, -0.001025291276164353, 0.004914933815598488, -0.09858298301696777, ...
DyRep: Learning Representations over Dynamic Graphs
https://openreview.net/forum?id=HyePrhR5KX
[ "Rakshit Trivedi", "Mehrdad Farajtabar", "Prasenjeet Biswal", "Hongyuan Zha" ]
Poster
null
Representation Learning over graph structured data has received significant attention recently due to its ubiquitous applicability. However, most advancements have been made in static graph settings while efforts for jointly learning dynamic of the graph and dynamic on the graph are still in an infant stage. Two fundam...
[ "Dynamic Graphs", "Representation Learning", "Dynamic Processes", "Temporal Point Process", "Attention", "Latent Representation" ]
Models Representation Learning over dynamic graphs as latent hidden process bridging two observed processes of Topological Evolution of and Interactions on dynamic graphs.
1,553
null
null
[ -0.016020027920603752, -0.013514269143342972, -0.003965671174228191, 0.04266967624425888, 0.030169157311320305, 0.017060941085219383, 0.024990567937493324, 0.02033553645014763, -0.01676897704601288, -0.03884540870785713, 0.047894541174173355, -0.024185314774513245, -0.06939507275819778, 0....
Label super-resolution networks
https://openreview.net/forum?id=rkxwShA9Ym
[ "Kolya Malkin", "Caleb Robinson", "Le Hou", "Rachel Soobitsky", "Jacob Czawlytko", "Dimitris Samaras", "Joel Saltz", "Lucas Joppa", "Nebojsa Jojic" ]
Poster
null
We present a deep learning-based method for super-resolving coarse (low-resolution) labels assigned to groups of image pixels into pixel-level (high-resolution) labels, given the joint distribution between those low- and high-resolution labels. This method involves a novel loss function that minimizes the distance betw...
[ "weakly supervised segmentation", "land cover mapping", "medical imaging" ]
Super-resolving coarse labels into pixel-level labels, applied to aerial imagery and medical scans.
1,552
null
null
[ -0.03058401122689247, -0.009797396138310432, -0.008801807649433613, 0.023975955322384834, 0.04687165468931198, 0.018259592354297638, 0.005479535553604364, -0.024961333721876144, -0.020294081419706345, -0.0412951223552227, -0.011619582772254944, 0.0016695214435458183, -0.04805901274085045, ...
Multi-step Retriever-Reader Interaction for Scalable Open-domain Question Answering
https://openreview.net/forum?id=HkfPSh05K7
[ "Rajarshi Das", "Shehzaad Dhuliawala", "Manzil Zaheer", "Andrew McCallum" ]
Poster
null
This paper introduces a new framework for open-domain question answering in which the retriever and the reader \emph{iteratively interact} with each other. The framework is agnostic to the architecture of the machine reading model provided it has \emph{access} to the token-level hidden representations of the reader. Th...
[ "Open domain Question Answering", "Reinforcement Learning", "Query reformulation" ]
Paragraph retriever and machine reader interacts with each other via reinforcement learning to yield large improvements on open domain datasets
1,551
1905.05733
title_snapshot
[ -0.03432149812579155, -0.06555280834436417, 0.01172560639679432, 0.05602841079235077, 0.03236961364746094, -0.00678950734436512, 0.017359580844640732, -0.009872307069599628, 0.007302420679479837, -0.002293238416314125, -0.012569512240588665, 0.043830495327711105, -0.05857913941144943, -0.0...
Capsule Graph Neural Network
https://openreview.net/forum?id=Byl8BnRcYm
[ "Zhang Xinyi", "Lihui Chen" ]
Poster
null
The high-quality node embeddings learned from the Graph Neural Networks (GNNs) have been applied to a wide range of node-based applications and some of them have achieved state-of-the-art (SOTA) performance. However, when applying node embeddings learned from GNNs to generate graph embeddings, the scalar node represent...
[ "CapsNet", "Graph embedding", "GNN" ]
Inspired by CapsNet, we propose a novel architecture for graph embeddings on the basis of node features extracted from GNN.
1,544
null
null
[ 0.008679721504449844, -0.03182566165924072, 0.015853028744459152, 0.03985704109072685, 0.023266293108463287, 0.014876622706651688, 0.0059364670887589455, 0.011246014386415482, -0.010759346187114716, -0.043684788048267365, 0.019283467903733253, -0.017705190926790237, -0.06304853409528732, 0...
Learning a Meta-Solver for Syntax-Guided Program Synthesis
https://openreview.net/forum?id=Syl8Sn0cK7
[ "Xujie Si", "Yuan Yang", "Hanjun Dai", "Mayur Naik", "Le Song" ]
Poster
null
We study a general formulation of program synthesis called syntax-guided synthesis(SyGuS) that concerns synthesizing a program that follows a given grammar and satisfies a given logical specification. Both the logical specification and the grammar have complex structures and can vary from task to task, posing significa...
[ "Syntax-guided Synthesis", "Context Free Grammar", "Logical Specification", "Representation Learning", "Meta Learning", "Reinforcement Learning" ]
We propose a meta-learning framework that learns a transferable policy from only weak supervision to solve synthesis tasks with different logical specifications and grammars.
1,543
null
null
[ -0.013572359457612038, -0.0031096357852220535, -0.02572775073349476, 0.036198072135448456, 0.04545699059963226, 0.04547741636633873, 0.030860476195812225, -0.008537562564015388, -0.02759794145822525, -0.020027849823236465, -0.0010836331639438868, 0.0377325601875782, -0.06642884761095047, -...
Stochastic Optimization of Sorting Networks via Continuous Relaxations
https://openreview.net/forum?id=H1eSS3CcKX
[ "Aditya Grover", "Eric Wang", "Aaron Zweig", "Stefano Ermon" ]
Poster
null
Sorting input objects is an important step in many machine learning pipelines. However, the sorting operator is non-differentiable with respect to its inputs, which prohibits end-to-end gradient-based optimization. In this work, we propose NeuralSort, a general-purpose continuous relaxation of the output of the sorting...
[ "continuous relaxations", "sorting", "permutation", "stochastic computation graphs", "Plackett-Luce" ]
We provide a continuous relaxation to the sorting operator, enabling end-to-end, gradient-based stochastic optimization.
1,541
1903.08850
title_snapshot
[ -0.04908856749534607, -0.02598000317811966, -0.009497452527284622, 0.047759391367435455, 0.0015489580109715462, 0.05151854082942009, 0.00008777665061643347, 0.010882037691771984, -0.03816669061779976, -0.05247141420841217, 0.005023143254220486, -0.021207084879279137, -0.05510375648736954, ...
Energy-Constrained Compression for Deep Neural Networks via Weighted Sparse Projection and Layer Input Masking
https://openreview.net/forum?id=BylBr3C9K7
[ "Haichuan Yang", "Yuhao Zhu", "Ji Liu" ]
Poster
null
Deep Neural Networks (DNNs) are increasingly deployed in highly energy-constrained environments such as autonomous drones and wearable devices while at the same time must operate in real-time. Therefore, reducing the energy consumption has become a major design consideration in DNN training. This paper proposes the fir...
[ "model compression", "inference energy saving", "deep neural network pruning" ]
null
1,540
1806.04321
title_snapshot
[ 0.016734834760427475, -0.028105895966291428, -0.027522053569555283, 0.03376628831028938, 0.05679125711321831, 0.06789594143629074, 0.01440412923693657, -0.01607491821050644, -0.03668860346078873, -0.04207572713494301, -0.017161240801215172, -0.014702738262712955, -0.059255875647068024, -0....
Composing Complex Skills by Learning Transition Policies
https://openreview.net/forum?id=rygrBhC5tQ
[ "Youngwoon Lee*", "Shao-Hua Sun*", "Sriram Somasundaram", "Edward S. Hu", "Joseph J. Lim" ]
Poster
null
Humans acquire complex skills by exploiting previously learned skills and making transitions between them. To empower machines with this ability, we propose a method that can learn transition policies which effectively connect primitive skills to perform sequential tasks without handcrafted rewards. To efficiently trai...
[ "reinforcement learning", "hierarchical reinforcement learning", "continuous control", "modular framework" ]
Transition policies enable agents to compose complex skills by smoothly connecting previously acquired primitive skills.
1,537
null
null
[ -0.024004239588975906, -0.031351905316114426, -0.01820707507431507, 0.04070500656962395, 0.04466864839196205, 0.012648308649659157, 0.011614290997385979, -0.02560696005821228, -0.029533328488469124, -0.028833512216806412, -0.04498277232050896, -0.009946408681571484, -0.051139168441295624, ...
Revealing interpretable object representations from human behavior
https://openreview.net/forum?id=ryxSrhC9KX
[ "Charles Y. Zheng", "Francisco Pereira", "Chris I. Baker", "Martin N. Hebart" ]
Poster
null
To study how mental object representations are related to behavior, we estimated sparse, non-negative representations of objects using human behavioral judgments on images representative of 1,854 object categories. These representations predicted a latent similarity structure between objects, which captured most of the...
[ "category representation", "sparse coding", "representation learning", "interpretable representations" ]
Human behavioral judgments are used to obtain sparse and interpretable representations of objects that generalize to other tasks
1,536
1901.02915
title_snapshot
[ -0.00853613018989563, 0.024838827550411224, -0.013367393985390663, 0.04061117768287659, 0.04216062277555466, 0.016549577936530113, 0.039507534354925156, 0.012727851048111916, -0.025646550580859184, -0.034433573484420776, -0.04397042840719223, -0.0002976526156999171, -0.06369686871767044, 0...
ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware
https://openreview.net/forum?id=HylVB3AqYm
[ "Han Cai", "Ligeng Zhu", "Song Han" ]
Poster
null
Neural architecture search (NAS) has a great impact by automatically designing effective neural network architectures. However, the prohibitive computational demand of conventional NAS algorithms (e.g. 10 4 GPU hours) makes it difficult to directly search the architectures on large-scale tasks (e.g. ImageNet). Differen...
[ "Neural Architecture Search", "Efficient Neural Networks" ]
Proxy-less neural architecture search for directly learning architectures on large-scale target task (ImageNet) while reducing the cost to the same level of normal training.
1,534
1812.00332
title_snapshot
[ 0.003405066207051277, -0.020266368985176086, -0.006441347301006317, 0.023071283474564552, 0.04286019876599312, 0.018078479915857315, 0.008771470747888088, 0.017391109839081764, -0.025751568377017975, -0.03929953649640083, 0.01986103318631649, -0.011133630760014057, -0.043993789702653885, 0...
A Generative Model For Electron Paths
https://openreview.net/forum?id=r1x4BnCqKX
[ "John Bradshaw", "Matt J. Kusner", "Brooks Paige", "Marwin H. S. Segler", "José Miguel Hernández-Lobato" ]
Poster
null
Chemical reactions can be described as the stepwise redistribution of electrons in molecules. As such, reactions are often depicted using "arrow-pushing" diagrams which show this movement as a sequence of arrows. We propose an electron path prediction model (ELECTRO) to learn these sequences directly from raw reaction ...
[ "Molecules", "Reaction Prediction", "Graph Neural Networks", "Deep Generative Models" ]
A generative model for reaction prediction that learns the mechanistic electron steps of a reaction directly from raw reaction data.
1,533
1805.10970
title_snapshot
[ -0.00910372007638216, 0.01808750256896019, -0.01841409131884575, 0.06175179034471512, 0.03675568848848343, -0.029360268265008926, -0.0023488544393330812, 0.009761625900864601, 0.015817036852240562, -0.03925525024533272, 0.01287752017378807, 0.024421386420726776, -0.056843530386686325, 0.00...
Learning to Infer and Execute 3D Shape Programs
https://openreview.net/forum?id=rylNH20qFQ
[ "Yonglong Tian", "Andrew Luo", "Xingyuan Sun", "Kevin Ellis", "William T. Freeman", "Joshua B. Tenenbaum", "Jiajun Wu" ]
Poster
null
Human perception of 3D shapes goes beyond reconstructing them as a set of points or a composition of geometric primitives: we also effortlessly understand higher-level shape structure such as the repetition and reflective symmetry of object parts. In contrast, recent advances in 3D shape sensing focus more on low-level...
[ "Program Synthesis", "3D Shape Modeling", "Self-supervised Learning" ]
We propose 3D shape programs, a structured, compositional shape representation. Our model learns to infer and execute shape programs to explain 3D shapes.
1,532
1901.02875
title_snapshot
[ 0.016404207795858383, 0.012172267772257328, -0.020078882575035095, 0.031403299421072006, 0.03228330239653587, 0.039817798882722855, 0.005872612819075584, 0.01104111596941948, -0.02087518386542797, -0.06760299205780029, -0.026963595300912857, 0.01687627099454403, -0.06325194239616394, -0.02...
Music Transformer: Generating Music with Long-Term Structure
https://openreview.net/forum?id=rJe4ShAcF7
[ "Cheng-Zhi Anna Huang", "Ashish Vaswani", "Jakob Uszkoreit", "Ian Simon", "Curtis Hawthorne", "Noam Shazeer", "Andrew M. Dai", "Matthew D. Hoffman", "Monica Dinculescu", "Douglas Eck" ]
Poster
null
Music relies heavily on repetition to build structure and meaning. Self-reference occurs on multiple timescales, from motifs to phrases to reusing of entire sections of music, such as in pieces with ABA structure. The Transformer (Vaswani et al., 2017), a sequence model based on self-attention, has achieved compellin...
[ "music generation" ]
We show the first successful use of Transformer in generating music that exhibits long-term structure.
1,531
null
null
[ -0.012744196690618992, -0.017616670578718185, -0.0016128438292071223, 0.014853239059448242, 0.034283820539712906, 0.0069905733689665794, 0.03881377354264259, 0.010585666634142399, -0.03673088178038597, -0.03034183755517006, 0.00015168719983194023, 0.013851399533450603, -0.04703967645764351, ...
Modeling the Long Term Future in Model-Based Reinforcement Learning
https://openreview.net/forum?id=SkgQBn0cF7
[ "Nan Rosemary Ke", "Amanpreet Singh", "Ahmed Touati", "Anirudh Goyal", "Yoshua Bengio", "Devi Parikh", "Dhruv Batra" ]
Poster
null
In model-based reinforcement learning, the agent interleaves between model learning and planning. These two components are inextricably intertwined. If the model is not able to provide sensible long-term prediction, the executed planer would exploit model flaws, which can yield catastrophic failures. This paper focus...
[ "model-based reinforcement learning", "variation inference" ]
incorporating, in the model, latent variables that encode future content improves the long-term prediction accuracy, which is critical for better planning in model-based RL.
1,527
null
null
[ -0.013209995813667774, -0.029657907783985138, -0.007304910570383072, 0.029128611087799072, 0.06191067397594452, 0.03872617706656456, -0.009127468802034855, 0.015608178451657295, -0.045806702226400375, -0.037300825119018555, -0.022732708603143692, 0.010590783320367336, -0.0545259565114975, ...
Model-Predictive Policy Learning with Uncertainty Regularization for Driving in Dense Traffic
https://openreview.net/forum?id=HygQBn0cYm
[ "Mikael Henaff", "Alfredo Canziani", "Yann LeCun" ]
Poster
null
Learning a policy using only observational data is challenging because the distribution of states it induces at execution time may differ from the distribution observed during training. In this work, we propose to train a policy while explicitly penalizing the mismatch between these two distributions over a fixed tim...
[ "model-based reinforcement learning", "stochastic video prediction", "autonomous driving" ]
A model-based RL approach which uses a differentiable uncertainty penalty to learn driving policies from purely observational data.
1,525
1901.02705
title_snapshot
[ 0.005908135790377855, -0.020244162529706955, -0.003964092116802931, 0.07352932542562485, 0.04181142896413803, 0.032738108187913895, 0.032623160630464554, 0.0019388437503948808, -0.015084505081176758, -0.052363522350788116, 0.0010497232433408499, 0.017406944185495377, -0.0565665066242218, -...
Learning to Screen for Fast Softmax Inference on Large Vocabulary Neural Networks
https://openreview.net/forum?id=ByeMB3Act7
[ "Patrick Chen", "Si Si", "Sanjiv Kumar", "Yang Li", "Cho-Jui Hsieh" ]
Poster
null
Neural language models have been widely used in various NLP tasks, including machine translation, next word prediction and conversational agents. However, it is challenging to deploy these models on mobile devices due to their slow prediction speed, where the bottleneck is to compute top candidates in the softmax layer...
[ "fast inference", "softmax computation", "natural language processing" ]
null
1,524
1810.12406
title_snapshot
[ -0.04245692864060402, -0.024097099900245667, 0.02938573993742466, 0.0199446901679039, 0.03868211433291435, 0.00820118747651577, 0.015062814578413963, 0.004545304458588362, -0.013577422127127647, -0.021468987688422203, -0.027987031266093254, 0.02934095822274685, -0.05279034376144409, 0.0196...
Interpolation-Prediction Networks for Irregularly Sampled Time Series
https://openreview.net/forum?id=r1efr3C9Ym
[ "Satya Narayan Shukla", "Benjamin Marlin" ]
Poster
null
In this paper, we present a new deep learning architecture for addressing the problem of supervised learning with sparse and irregularly sampled multivariate time series. The architecture is based on the use of a semi-parametric interpolation network followed by the application of a prediction network. The interpolatio...
[ "irregular sampling", "multivariate time series", "supervised learning", "interpolation", "missing data" ]
This paper presents a new deep learning architecture for addressing the problem of supervised learning with sparse and irregularly sampled multivariate time series.
1,523
1909.07782
title_snapshot
[ -0.000857489590998739, -0.024264156818389893, -0.017110755667090416, 0.02947148308157921, 0.06518756598234177, 0.05156201124191284, 0.037661805748939514, -0.016795748844742775, -0.016620907932519913, -0.06467773765325546, 0.04514202103018761, -0.01206095702946186, -0.04317880794405937, 0.0...
Contingency-Aware Exploration in Reinforcement Learning
https://openreview.net/forum?id=HyxGB2AcY7
[ "Jongwook Choi", "Yijie Guo", "Marcin Moczulski", "Junhyuk Oh", "Neal Wu", "Mohammad Norouzi", "Honglak Lee" ]
Poster
null
This paper investigates whether learning contingency-awareness and controllable aspects of an environment can lead to better exploration in reinforcement learning. To investigate this question, we consider an instantiation of this hypothesis evaluated on the Arcade Learning Element (ALE). In this study, we develop an a...
[ "Reinforcement Learning", "Exploration", "Contingency-Awareness" ]
We investigate contingency-awareness and controllable aspects in exploration and achieve state-of-the-art performance on Montezuma's Revenge without expert demonstrations.
1,520
1811.01483
title_snapshot
[ -0.02900741435587406, -0.01236736960709095, -0.027607688680291176, 0.048508577048778534, 0.036149486899375916, 0.0006749709136784077, 0.040670789778232574, 0.01255074329674244, -0.05342820659279823, -0.026325849816203117, -0.01486158836632967, 0.006121616344898939, -0.041105467826128006, -...
Neural Graph Evolution: Towards Efficient Automatic Robot Design
https://openreview.net/forum?id=BkgWHnR5tm
[ "Tingwu Wang", "Yuhao Zhou", "Sanja Fidler", "Jimmy Ba" ]
Poster
null
Despite the recent successes in robotic locomotion control, the design of robot relies heavily on human engineering. Automatic robot design has been a long studied subject, but the recent progress has been slowed due to the large combinatorial search space and the difficulty in evaluating the found candidates. To addre...
[ "Reinforcement learning", "graph neural networks", "robotics", "deep learning", "transfer learning" ]
Automatic robotic design search with graph neural networks
1,518
1906.05370
title_snapshot
[ -0.027342908084392548, -0.013678055256605148, -0.032269977033138275, 0.04952958971261978, 0.058454595506191254, 0.042143795639276505, 0.014526737853884697, 0.021879049018025398, -0.036041080951690674, -0.07612530887126923, 0.018922824412584305, -0.006615887861698866, -0.060243330895900726, ...
Selfless Sequential Learning
https://openreview.net/forum?id=Bkxbrn0cYX
[ "Rahaf Aljundi", "Marcus Rohrbach", "Tinne Tuytelaars" ]
Poster
null
Sequential learning, also called lifelong learning, studies the problem of learning tasks in a sequence with access restricted to only the data of the current task. In this paper we look at a scenario with fixed model capacity, and postulate that the learning process should not be selfish, i.e. it should account for fu...
[ "Lifelong learning", "Continual Learning", "Sequential learning", "Regularization" ]
A regularization strategy for improving the performance of sequential learning
1,517
1806.05421
title_snapshot
[ -0.006211030762642622, -0.02332957647740841, -0.005594572983682156, 0.025709044188261032, 0.05180305242538452, 0.02477922849357128, 0.019286328926682472, 0.00671537546440959, -0.05917300283908844, -0.027462376281619072, 0.0008213849505409598, -0.00022160417574923486, -0.03905511274933815, ...
Learning Particle Dynamics for Manipulating Rigid Bodies, Deformable Objects, and Fluids
https://openreview.net/forum?id=rJgbSn09Ym
[ "Yunzhu Li", "Jiajun Wu", "Russ Tedrake", "Joshua B. Tenenbaum", "Antonio Torralba" ]
Poster
null
Real-life control tasks involve matters of various substances---rigid or soft bodies, liquid, gas---each with distinct physical behaviors. This poses challenges to traditional rigid-body physics engines. Particle-based simulators have been developed to model the dynamics of these complex scenes; however, relying on app...
[ "Dynamics modeling", "Control", "Particle-Based Representation" ]
Learning particle dynamics with dynamic interaction graphs for simulating and control rigid bodies, deformable objects, and fluids.
1,513
1810.01566
title_snapshot
[ -0.02545994333922863, 0.013929275795817375, 0.006622659042477608, 0.0264179278165102, 0.03332561254501343, 0.029584502801299095, -0.013553475961089134, 0.019815051928162575, -0.06506094336509705, -0.05341000482439995, -0.00008742991485632956, -0.021316519007086754, -0.05477997660636902, 0....
Representing Formal Languages: A Comparison Between Finite Automata and Recurrent Neural Networks
https://openreview.net/forum?id=H1zeHnA9KX
[ "Joshua J. Michalenko", "Ameesh Shah", "Abhinav Verma", "Richard G. Baraniuk", "Swarat Chaudhuri", "Ankit B. Patel" ]
Poster
null
We investigate the internal representations that a recurrent neural network (RNN) uses while learning to recognize a regular formal language. Specifically, we train a RNN on positive and negative examples from a regular language, and ask if there is a simple decoding function that maps states of this RNN to states of t...
[ "Language recognition", "Recurrent Neural Networks", "Representation Learning", "deterministic finite automaton", "automaton" ]
Finite Automata Can be Linearly decoded from Language-Recognizing RNNs using low coarseness abstraction functions and high accuracy decoders.
1,511
1902.10297
title_snapshot
[ -0.036113884299993515, -0.009396553970873356, -0.033682700246572495, 0.04298475384712219, 0.051704131066799164, 0.04634387418627739, 0.02954256907105446, 0.021151045337319374, -0.04358820617198944, -0.016123704612255096, -0.0057191974483430386, 0.020099660381674767, -0.07260517030954361, 0...
Disjoint Mapping Network for Cross-modal Matching of Voices and Faces
https://openreview.net/forum?id=B1exrnCcF7
[ "Yandong Wen", "Mahmoud Al Ismail", "Weiyang Liu", "Bhiksha Raj", "Rita Singh" ]
Poster
null
We propose a novel framework, called Disjoint Mapping Network (DIMNet), for cross-modal biometric matching, in particular of voices and faces. Different from the existing methods, DIMNet does not explicitly learn the joint relationship between the modalities. Instead, DIMNet learns a shared representation for different...
[ "cross-modal matching", "voices", "faces" ]
null
1,510
1807.04836
title_snapshot
[ -0.007384192664176226, -0.024956464767456055, 0.003196344245225191, 0.04053773730993271, 0.023344673216342926, 0.06111317500472069, 0.05865279585123062, -0.0070066810585558414, -0.025278905406594276, -0.0450587160885334, -0.010811947286128998, 0.013829275965690613, -0.0922793373465538, -0....
Learning Procedural Abstractions and Evaluating Discrete Latent Temporal Structure
https://openreview.net/forum?id=ByleB2CcKm
[ "Karan Goel", "Emma Brunskill" ]
Poster
null
Clustering methods and latent variable models are often used as tools for pattern mining and discovery of latent structure in time-series data. In this work, we consider the problem of learning procedural abstractions from possibly high-dimensional observational sequences, such as video demonstrations. Given a dataset ...
[ "learning procedural abstractions", "latent variable modeling", "evaluation criteria" ]
null
1,508
null
null
[ 0.009453019127249718, -0.03722713887691498, -0.012625099159777164, 0.04649723321199417, 0.04196639358997345, 0.012294139713048935, 0.049155548214912415, 0.004620982334017754, -0.035887766629457474, -0.039300236850976944, -0.0004071770526934415, -0.003119029337540269, -0.03637765720486641, ...
Learning to Design RNA
https://openreview.net/forum?id=ByfyHh05tQ
[ "Frederic Runge", "Danny Stoll", "Stefan Falkner", "Frank Hutter" ]
Poster
null
Designing RNA molecules has garnered recent interest in medicine, synthetic biology, biotechnology and bioinformatics since many functional RNA molecules were shown to be involved in regulatory processes for transcription, epigenetics and translation. Since an RNA's function depends on its structural properties, the RN...
[ "matter engineering", "bioinformatics", "rna design", "reinforcement learning", "meta learning", "neural architecture search", "hyperparameter optimization" ]
We learn to solve the RNA Design problem with reinforcement learning using meta learning and autoML approaches.
1,504
1812.11951
title_snapshot
[ 0.01870989054441452, -0.0036404479760676622, -0.016029255464673042, 0.025560470297932625, 0.04410650581121445, -0.0012807552702724934, 0.04249424487352371, 0.016367655247449875, 0.016436802223324776, -0.030812103301286697, 0.02539580501616001, 0.03660685196518898, -0.059202875941991806, 0....
Cost-Sensitive Robustness against Adversarial Examples
https://openreview.net/forum?id=BygANhA9tQ
[ "Xiao Zhang", "David Evans" ]
Poster
null
Several recent works have developed methods for training classifiers that are certifiably robust against norm-bounded adversarial perturbations. These methods assume that all the adversarial transformations are equally important, which is seldom the case in real-world applications. We advocate for cost-sensitive robust...
[ "Certified robustness", "Adversarial examples", "Cost-sensitive learning" ]
A general method for training certified cost-sensitive robust classifier against adversarial perturbations
1,496
1810.09225
title_snapshot
[ -0.0077033257111907005, -0.019264012575149536, -0.0021761013194918633, 0.029496680945158005, 0.046325329691171646, 0.028888339176774025, 0.024222278967499733, -0.027555879205465317, 0.00713697075843811, -0.041512563824653625, -0.007633083965629339, -0.002187746111303568, -0.0658932477235794,...
Combinatorial Attacks on Binarized Neural Networks
https://openreview.net/forum?id=S1lTEh09FQ
[ "Elias B Khalil", "Amrita Gupta", "Bistra Dilkina" ]
Poster
null
Binarized Neural Networks (BNNs) have recently attracted significant interest due to their computational efficiency. Concurrently, it has been shown that neural networks may be overly sensitive to ``attacks" -- tiny adversarial changes in the input -- which may be detrimental to their use in safety-critical domains. De...
[ "binarized neural networks", "combinatorial optimization", "integer programming" ]
Gradient-based attacks on binarized neural networks are not effective due to the non-differentiability of such networks; Our IPROP algorithm solves this problem using integer optimization
1,495
1810.03538
title_snapshot
[ -0.006788248661905527, -0.011681891977787018, -0.02701864391565323, 0.05611111596226692, 0.021723635494709015, 0.02184276282787323, 0.017739562317728996, -0.031211666762828827, -0.0423126295208931, -0.033218078315258026, 0.013325875625014305, -0.00048451012116856873, -0.047248177230358124, ...
A Variational Inequality Perspective on Generative Adversarial Networks
https://openreview.net/forum?id=r1laEnA5Ym
[ "Gauthier Gidel", "Hugo Berard", "Gaëtan Vignoud", "Pascal Vincent", "Simon Lacoste-Julien" ]
Poster
null
Generative adversarial networks (GANs) form a generative modeling approach known for producing appealing samples, but they are notably difficult to train. One common way to tackle this issue has been to propose new formulations of the GAN objective. Yet, surprisingly few studies have looked at optimization methods desi...
[ "optimization", "variational inequality", "games", "saddle point", "extrapolation", "averaging", "extragradient", "generative modeling", "generative adversarial network" ]
We cast GANs in the variational inequality framework and import techniques from this literature to optimize GANs better; we give algorithmic extensions and empirically test their performance for training GANs.
1,491
1802.10551
title_snapshot
[ -0.02293345518410206, 0.004530074540525675, -0.025215407833456993, 0.06460070610046387, 0.010495616123080254, 0.053866442292928696, 0.016292555257678032, -0.0051912907510995865, -0.010929057374596596, -0.04206092655658722, -0.036713436245918274, 0.00029414737946353853, -0.07830926775932312, ...
Multiple-Attribute Text Rewriting
https://openreview.net/forum?id=H1g2NhC5KQ
[ "Guillaume Lample", "Sandeep Subramanian", "Eric Smith", "Ludovic Denoyer", "Marc'Aurelio Ranzato", "Y-Lan Boureau" ]
Poster
null
The dominant approach to unsupervised "style transfer'' in text is based on the idea of learning a latent representation, which is independent of the attributes specifying its "style''. In this paper, we show that this condition is not necessary and is not always met in practice, even with domain adversarial training t...
[ "controllable text generation", "generative models", "conditional generative models", "style transfer" ]
A system for rewriting text conditioned on multiple controllable attributes
1,489
null
null
[ 0.015303513966500759, -0.022587312385439873, -0.012990349903702736, 0.07245560735464096, 0.05063601955771446, 0.005611439701169729, 0.008883470669388771, 0.007822899147868156, 0.015746159479022026, -0.026475287973880768, -0.045451901853084564, 0.04891789332032204, -0.06925439834594727, -0....
Multi-Agent Dual Learning
https://openreview.net/forum?id=HyGhN2A5tm
[ "Yiren Wang", "Yingce Xia", "Tianyu He", "Fei Tian", "Tao Qin", "ChengXiang Zhai", "Tie-Yan Liu" ]
Poster
null
Dual learning has attracted much attention in machine learning, computer vision and natural language processing communities. The core idea of dual learning is to leverage the duality between the primal task (mapping from domain X to domain Y) and dual task (mapping from domain Y to X) to boost the performances of both ...
[ "Dual Learning", "Machine Learning", "Neural Machine Translation" ]
null
1,486
null
null
[ -0.005312667693942785, -0.018491234630346298, 0.0028504873625934124, 0.005786978639662266, 0.028479106724262238, 0.03491317853331566, 0.03975332900881767, 0.013923984952270985, 0.005318932235240936, -0.03664231672883034, -0.005589631851762533, 0.03960518166422844, -0.04418882355093956, -0....
Learning sparse relational transition models
https://openreview.net/forum?id=SJxsV2R5FQ
[ "Victoria Xia", "Zi Wang", "Kelsey Allen", "Tom Silver", "Leslie Pack Kaelbling" ]
Poster
null
We present a representation for describing transition models in complex uncertain domains using relational rules. For any action, a rule selects a set of relevant objects and computes a distribution over properties of just those objects in the resulting state given their properties in the previous state. An iterative...
[ "Deictic reference", "relational model", "rule-based transition model" ]
A new approach that learns a representation for describing transition models in complex uncertaindomains using relational rules.
1,481
1810.11177
title_snapshot
[ -0.02000042423605919, 0.01241046842187643, -0.018216580152511597, 0.029808519408106804, 0.05083063244819641, 0.004838856868445873, -0.01658623293042183, 0.012591435573995113, -0.03230377286672592, -0.015946200117468834, -0.007767405826598406, 0.030349591746926308, -0.04629902169108391, 0.0...
SPIGAN: Privileged Adversarial Learning from Simulation
https://openreview.net/forum?id=rkxoNnC5FQ
[ "Kuan-Hui Lee", "German Ros", "Jie Li", "Adrien Gaidon" ]
Poster
null
Deep Learning for Computer Vision depends mainly on the source of supervision. Photo-realistic simulators can generate large-scale automatically labeled synthetic data, but introduce a domain gap negatively impacting performance. We propose a new unsupervised domain adaptation algorithm, called SPIGAN, relying on Simul...
[ "domain adaptation", "GAN", "semantic segmentation", "simulation", "privileged information" ]
An unsupervised sim-to-real domain adaptation method for semantic segmentation using privileged information from a simulator with GAN-based image translation.
1,479
1810.03756
title_snapshot
[ -0.01351255178451538, -0.02016124129295349, 0.009115992113947868, 0.03735966980457306, 0.025995057076215744, 0.019260156899690628, 0.015431074425578117, 0.0002706713567022234, -0.005346348043531179, -0.053987372666597366, -0.044416699558496475, -0.012642549350857735, -0.058979347348213196, ...
Universal Stagewise Learning for Non-Convex Problems with Convergence on Averaged Solutions
https://openreview.net/forum?id=Syx5V2CcFm
[ "Zaiyi Chen", "Zhuoning Yuan", "Jinfeng Yi", "Bowen Zhou", "Enhong Chen", "Tianbao Yang" ]
Poster
null
Although stochastic gradient descent (SGD) method and its variants (e.g., stochastic momentum methods, AdaGrad) are algorithms of choice for solving non-convex problems (especially deep learning), big gaps still remain between the theory and the practice with many questions unresolved. For example, there is still a la...
[ "optimization", "sgd", "adagrad" ]
null
1,475
1808.06296
title_snapshot
[ -0.0549156591296196, -0.040362898260354996, 0.015093074180185795, 0.04064884036779404, 0.029303817078471184, 0.03524279221892357, 0.051953136920928955, 0.011058746837079525, -0.02550729550421238, -0.03957397863268852, -0.0035529504530131817, -0.02616189233958721, -0.039581090211868286, 0.0...
Reasoning About Physical Interactions with Object-Oriented Prediction and Planning
https://openreview.net/forum?id=HJx9EhC9tQ
[ "Michael Janner", "Sergey Levine", "William T. Freeman", "Joshua B. Tenenbaum", "Chelsea Finn", "Jiajun Wu" ]
Poster
null
Object-based factorizations provide a useful level of abstraction for interacting with the world. Building explicit object representations, however, often requires supervisory signals that are difficult to obtain in practice. We present a paradigm for learning object-centric representations for physical scene understan...
[ "structured scene representation", "predictive models", "intuitive physics", "self-supervised learning" ]
We present a framework for learning object-centric representations suitable for planning in tasks that require an understanding of physics.
1,474
1812.10972
title_snapshot
[ -0.014636178500950336, 0.011038209311664104, -0.007998557761311531, 0.007836190052330494, 0.03507472947239876, 0.0029353888239711523, -0.0013865007786080241, -0.0038283602334558964, -0.036788925528526306, -0.017499133944511414, -0.04370226338505745, 0.002991366432979703, -0.07407958805561066...
Posterior Attention Models for Sequence to Sequence Learning
https://openreview.net/forum?id=BkltNhC9FX
[ "Shiv Shankar", "Sunita Sarawagi" ]
Poster
null
Modern neural architectures critically rely on attention for mapping structured inputs to sequences. In this paper we show that prevalent attention architectures do not adequately model the dependence among the attention and output tokens across a predicted sequence. We present an alternative architecture called Poste...
[ "posterior inference", "attention", "seq2seq learning", "translation" ]
Computing attention based on posterior distribution leads to more meaningful attention and better performance
1,473
null
null
[ 0.013247719965875149, 0.021596508100628853, -0.01778341457247734, 0.025441523641347885, 0.006156215909868479, 0.056110695004463196, 0.043064165860414505, 0.030497098341584206, -0.022394252941012383, -0.007727081421762705, -0.008854077197611332, 0.028991950675845146, -0.05617706850171089, -...
NOODL: Provable Online Dictionary Learning and Sparse Coding
https://openreview.net/forum?id=HJeu43ActQ
[ "Sirisha Rambhatla", "Xingguo Li", "Jarvis Haupt" ]
Poster
null
We consider the dictionary learning problem, where the aim is to model the given data as a linear combination of a few columns of a matrix known as a dictionary, where the sparse weights forming the linear combination are known as coefficients. Since the dictionary and coefficients, parameterizing the linear model are ...
[ "dictionary learning", "provable dictionary learning", "online dictionary learning", "sparse coding", "support recovery", "iterative hard thresholding", "matrix factorization", "neural architectures", "neural networks", "noodl" ]
We present a provable algorithm for exactly recovering both factors of the dictionary learning model.
1,464
1902.11261
title_snapshot
[ -0.035959821194410324, 0.000392153364373371, 0.02306460589170456, 0.023439425975084305, 0.04086671769618988, 0.024203049018979073, 0.013799927197396755, 0.028779128566384315, -0.07063206285238266, -0.029391884803771973, 0.001431416254490614, -0.003856024704873562, -0.06733038276433945, -0....
RelGAN: Relational Generative Adversarial Networks for Text Generation
https://openreview.net/forum?id=rJedV3R5tm
[ "Weili Nie", "Nina Narodytska", "Ankit Patel" ]
Poster
null
Generative adversarial networks (GANs) have achieved great success at generating realistic images. However, the text generation still remains a challenging task for modern GAN architectures. In this work, we propose RelGAN, a new GAN architecture for text generation, consisting of three main components: a relational me...
[ "RelGAN", "text generation", "relational memory", "Gumbel-Softmax relaxation", "multiple embedded representations" ]
null
1,462
null
null
[ -0.013116925954818726, -0.020314766094088554, 0.02220906689763069, 0.05168231949210167, 0.020475303754210472, -0.0021363648120313883, 0.003074676962569356, 0.023105256259441376, -0.015698177739977837, -0.028509188443422318, -0.030635617673397064, 0.015577197074890137, -0.07326801121234894, ...
Do Deep Generative Models Know What They Don't Know?
https://openreview.net/forum?id=H1xwNhCcYm
[ "Eric Nalisnick", "Akihiro Matsukawa", "Yee Whye Teh", "Dilan Gorur", "Balaji Lakshminarayanan" ]
Poster
null
A neural network deployed in the wild may be asked to make predictions for inputs that were drawn from a different distribution than that of the training data. A plethora of work has demonstrated that it is easy to find or synthesize inputs for which a neural network is highly confident yet wrong. Generative models a...
[ "deep generative models", "out-of-distribution inputs", "flow-based models", "uncertainty", "density" ]
null
1,461
1810.09136
title_snapshot
[ 0.011271880939602852, -0.02461191825568676, -0.008436650969088078, 0.058526478707790375, 0.034061990678310394, 0.031659673899412155, 0.02352520078420639, 0.005983612034469843, -0.021020878106355667, -0.05362443998456001, -0.02199671044945717, -0.012070504017174244, -0.053765181452035904, 0...
K for the Price of 1: Parameter-efficient Multi-task and Transfer Learning
https://openreview.net/forum?id=BJxvEh0cFQ
[ "Pramod Kaushik Mudrakarta", "Mark Sandler", "Andrey Zhmoginov", "Andrew Howard" ]
Poster
null
We introduce a novel method that enables parameter-efficient transfer and multi-task learning with deep neural networks. The basic approach is to learn a model patch - a small set of parameters - that will specialize to each task, instead of fine-tuning the last layer or the entire network. For instance, we show that l...
[ "deep learning", "mobile", "transfer learning", "multi-task learning", "computer vision", "small models", "imagenet", "inception", "batch normalization" ]
A novel and practically effective method to adapt pretrained neural networks to new tasks by retraining a minimal (e.g., less than 2%) number of parameters
1,458
1810.10703
title_snapshot
[ -0.0022277485113590956, -0.018161095678806305, 0.00821298360824585, 0.03211293742060661, 0.025461388751864433, 0.05577417090535164, 0.006754807662218809, -0.012913684360682964, 0.013187329284846783, -0.05947168543934822, -0.003416449297219515, 0.00261857477016747, -0.0676242858171463, -0.0...
MisGAN: Learning from Incomplete Data with Generative Adversarial Networks
https://openreview.net/forum?id=S1lDV3RcKm
[ "Steven Cheng-Xian Li", "Bo Jiang", "Benjamin Marlin" ]
Poster
null
Generative adversarial networks (GANs) have been shown to provide an effective way to model complex distributions and have obtained impressive results on various challenging tasks. However, typical GANs require fully-observed data during training. In this paper, we present a GAN-based framework for learning from comple...
[ "generative models", "missing data" ]
This paper presents a GAN-based framework for learning the distribution from high-dimensional incomplete data.
1,457
1902.09599
title_snapshot
[ -0.019628295674920082, -0.022252950817346573, -0.03739466518163681, 0.08115650713443756, 0.028393970802426338, 0.018514351919293404, 0.007147609256207943, -0.0009743983391672373, -0.0546974241733551, -0.057679906487464905, -0.024036245420575142, 0.013160870410501957, -0.057750154286623, 0....
Learnable Embedding Space for Efficient Neural Architecture Compression
https://openreview.net/forum?id=S1xLN3C9YX
[ "Shengcao Cao", "Xiaofang Wang", "Kris M. Kitani" ]
Poster
null
We propose a method to incrementally learn an embedding space over the domain of network architectures, to enable the careful selection of architectures for evaluation during compressed architecture search. Given a teacher network, we search for a compressed network architecture by using Bayesian Optimization (BO) with...
[ "Network Compression", "Neural Architecture Search", "Bayesian Optimization", "Architecture Embedding" ]
We propose a method to incrementally learn an embedding space over the domain of network architectures, to enable the careful selection of architectures for evaluation during compressed architecture search.
1,455
1902.00383
title_snapshot
[ 0.0002110770292347297, -0.02892248146235943, -0.020980024710297585, 0.04326115921139717, 0.05218667909502983, 0.05621463432908058, 0.010513614863157272, -0.00102997908834368, -0.0052070920355618, -0.04315780848264694, 0.007242199499160051, -0.008123827166855335, -0.04523292928934097, -0.02...
Guiding Policies with Language via Meta-Learning
https://openreview.net/forum?id=HkgSEnA5KQ
[ "John D. Co-Reyes", "Abhishek Gupta", "Suvansh Sanjeev", "Nick Altieri", "Jacob Andreas", "John DeNero", "Pieter Abbeel", "Sergey Levine" ]
Poster
null
Behavioral skills or policies for autonomous agents are conventionally learned from reward functions, via reinforcement learning, or from demonstrations, via imitation learning. However, both modes of task specification have their disadvantages: reward functions require manual engineering, while demonstrations require ...
[ "meta-learning", "language grounding", "interactive" ]
We propose a meta-learning method for interactively correcting policies with natural language.
1,449
1811.07882
title_snapshot
[ -0.032421305775642395, 0.0032808182295411825, -0.014126881025731564, 0.04070378094911575, 0.03290944918990135, 0.025597717612981796, 0.045939888805150986, 0.00025025176000781357, -0.040355756878852844, -0.00951279979199171, -0.044635288417339325, 0.05898843705654144, -0.06046760082244873, ...
Active Learning with Partial Feedback
https://openreview.net/forum?id=HJfSEnRqKQ
[ "Peiyun Hu", "Zachary C. Lipton", "Anima Anandkumar", "Deva Ramanan" ]
Poster
null
While many active learning papers assume that the learner can simply ask for a label and receive it, real annotation often presents a mismatch between the form of a label (say, one among many classes), and the form of an annotation (typically yes/no binary feedback). To annotate examples corpora for multiclass classifi...
[ "Active Learning", "Learning from Partial Feedback" ]
We provide a new perspective on training a machine learning model from scratch in hierarchical label setting, i.e. thinking of it as two-way communication between human and algorithms, and study how we can both measure and improve the efficiency.
1,448
1802.07427
title_snapshot
[ -0.011580915190279484, -0.05838824063539505, -0.021445047110319138, 0.05443089082837105, -0.0006786864250898361, 0.015370762906968594, -0.0024841828271746635, -0.01957538351416588, -0.01984117366373539, -0.01949218474328518, -0.03707514703273773, 0.019973844289779663, -0.05375305563211441, ...
On the Sensitivity of Adversarial Robustness to Input Data Distributions
https://openreview.net/forum?id=S1xNEhR9KX
[ "Gavin Weiguang Ding", "Kry Yik Chau Lui", "Xiaomeng Jin", "Luyu Wang", "Ruitong Huang" ]
Poster
null
Neural networks are vulnerable to small adversarial perturbations. Existing literature largely focused on understanding and mitigating the vulnerability of learned models. In this paper, we demonstrate an intriguing phenomenon about the most popular robust training method in the literature, adversarial training: Advers...
[ "adversarial robustness", "adversarial training", "PGD training", "adversarial perturbation", "input data distribution" ]
Robustness performance of PGD trained models are sensitive to semantics-preserving transformation of image datasets, which implies the trickiness of evaluation of robust learning algorithms in practice.
1,443
1902.08336
title_snapshot
[ -0.028090255334973335, -0.02158922515809536, 0.0051003070548176765, 0.05697259679436684, 0.04517922177910805, 0.011849228292703629, 0.029170548543334007, -0.0312756784260273, -0.010140610858798027, -0.030388645827770233, -0.01414306741207838, -0.015959639102220535, -0.07059735804796219, 0....
Efficient Multi-Objective Neural Architecture Search via Lamarckian Evolution
https://openreview.net/forum?id=ByME42AqK7
[ "Thomas Elsken", "Jan Hendrik Metzen", "Frank Hutter" ]
Poster
null
Architecture search aims at automatically finding neural architectures that are competitive with architectures designed by human experts. While recent approaches have achieved state-of-the-art predictive performance for image recognition, they are problematic under resource constraints for two reasons: (1) the neural a...
[ "Neural Architecture Search", "AutoML", "AutoDL", "Deep Learning", "Evolutionary Algorithms", "Multi-Objective Optimization" ]
We propose a method for efficient Multi-Objective Neural Architecture Search based on Lamarckian inheritance and evolutionary algorithms.
1,442
1804.09081
title_snapshot
[ -0.03526988998055458, 0.005545563064515591, -0.004219039808958769, 0.019635193049907684, 0.03176321089267731, 0.05784140154719353, 0.02349032834172249, 0.0357016995549202, -0.03741363435983658, -0.05121008679270744, 0.009804047644138336, -0.0249355249106884, -0.047176796942949295, -0.03252...
DISTRIBUTIONAL CONCAVITY REGULARIZATION FOR GANS
https://openreview.net/forum?id=SklEEnC5tQ
[ "Shoichiro Yamaguchi", "Masanori Koyama" ]
Poster
null
We propose Distributional Concavity (DC) regularization for Generative Adversarial Networks (GANs), a functional gradient-based method that promotes the entropy of the generator distribution and works against mode collapse. Our DC regularization is an easy-to-implement method that can be used in combination with the c...
[ "Generative Adversarial Networks", "regularization", "optimal transport", "functional gradient", "convex analysis" ]
null
1,441
null
null
[ 0.004958071280270815, -0.024581143632531166, -0.01136451493948698, 0.04840051382780075, 0.01911361888051033, 0.03034297376871109, 0.010923686437308788, -0.00860204640775919, -0.010319803841412067, -0.06238890439271927, -0.011669737286865711, -0.009587179869413376, -0.04925930127501488, 0.0...
Characterizing Audio Adversarial Examples Using Temporal Dependency
https://openreview.net/forum?id=r1g4E3C9t7
[ "Zhuolin Yang", "Bo Li", "Pin-Yu Chen", "Dawn Song" ]
Poster
null
Recent studies have highlighted adversarial examples as a ubiquitous threat to different neural network models and many downstream applications. Nonetheless, as unique data properties have inspired distinct and powerful learning principles, this paper aims to explore their potentials towards mitigating adversarial inp...
[ "audio adversarial example", "mitigation", "detection", "machine learning" ]
Adversarial audio discrimination using temporal dependency
1,439
1809.10875
title_snapshot
[ -0.008278397843241692, -0.012599927373230457, -0.018562184646725655, 0.05848819389939308, 0.03314795345067978, 0.01924370974302292, 0.055364470928907394, -0.01953769661486149, -0.018714457750320435, -0.030701691284775734, -0.0027383442502468824, 0.025181489065289497, -0.0520925335586071, 0...
GANSynth: Adversarial Neural Audio Synthesis
https://openreview.net/forum?id=H1xQVn09FX
[ "Jesse Engel", "Kumar Krishna Agrawal", "Shuo Chen", "Ishaan Gulrajani", "Chris Donahue", "Adam Roberts" ]
Poster
null
Efficient audio synthesis is an inherently difficult machine learning task, as human perception is sensitive to both global structure and fine-scale waveform coherence. Autoregressive models, such as WaveNet, model local structure at the expense of global latent structure and slow iterative sampling, while Generative A...
[ "GAN", "Audio", "WaveNet", "NSynth", "Music" ]
High-quality audio synthesis with GANs
1,437
1902.08710
title_snapshot
[ 0.0018936267588287592, -0.012691900134086609, -0.013436108827590942, 0.042571116238832474, 0.014350099489092827, 0.010643119923770428, 0.029356058686971664, -0.0014156034449115396, -0.021516963839530945, -0.06574555486440659, -0.004974950570613146, 0.010427931323647499, -0.041553471237421036...
Toward Understanding the Impact of Staleness in Distributed Machine Learning
https://openreview.net/forum?id=BylQV305YQ
[ "Wei Dai", "Yi Zhou", "Nanqing Dong", "Hao Zhang", "Eric Xing" ]
Poster
null
Most distributed machine learning (ML) systems store a copy of the model parameters locally on each machine to minimize network communication. In practice, in order to reduce synchronization waiting time, these copies of the model are not necessarily updated in lock-step, and can become stale. Despite much development ...
[]
Empirical and theoretical study of the effects of staleness in non-synchronous execution on machine learning algorithms.
1,436
1810.03264
title_snapshot
[ -0.03351148962974548, -0.03937739133834839, -0.010715884156525135, 0.07426925748586655, 0.05501927435398102, 0.0014971141936257482, 0.03214532136917114, 0.01121083740144968, -0.04051082208752632, -0.03848418965935707, 0.0022953771986067295, -0.0014976949896663427, -0.05437590554356575, 0.0...
Beyond Greedy Ranking: Slate Optimization via List-CVAE
https://openreview.net/forum?id=r1xX42R5Fm
[ "Ray Jiang", "Sven Gowal", "Yuqiu Qian", "Timothy Mann", "Danilo J. Rezende" ]
Poster
null
The conventional approach to solving the recommendation problem greedily ranks individual document candidates by prediction scores. However, this method fails to optimize the slate as a whole, and hence, often struggles to capture biases caused by the page layout and document interdepedencies. The slate recommendation ...
[ "CVAE", "VAE", "recommendation system", "slate optimization", "whole page optimization" ]
We used a CVAE type model structure to learn to directly generate slates/whole pages for recommendation systems.
1,434
1803.01682
title_snapshot
[ 0.015341839753091335, -0.002088972134515643, -0.02001689001917839, 0.0342111699283123, 0.03859415277838707, 0.00691597443073988, 0.001541280304081738, -0.013661730103194714, 0.0030883613508194685, -0.03701813519001007, -0.0350150503218174, 0.0013463386567309499, -0.058879390358924866, -0.0...
G-SGD: Optimizing ReLU Neural Networks in its Positively Scale-Invariant Space
https://openreview.net/forum?id=SyxfEn09Y7
[ "Qi Meng", "Shuxin Zheng", "Huishuai Zhang", "Wei Chen", "Qiwei Ye", "Zhi-Ming Ma", "Nenghai Yu", "Tie-Yan Liu" ]
Poster
null
It is well known that neural networks with rectified linear units (ReLU) activation functions are positively scale-invariant. Conventional algorithms like stochastic gradient descent optimize the neural networks in the vector space of weights, which is, however, not positively scale-invariant. This mismatch may lead to...
[ "optimization", "neural network", "irreducible positively scale-invariant space", "deep learning" ]
null
1,431
1802.03713
title_snapshot
[ -0.051851045340299606, -0.012184797786176205, 0.030467450618743896, 0.01518294122070074, 0.038466814905405045, 0.06154415011405945, 0.026618320494890213, 0.0095633864402771, -0.019564544782042503, -0.041778117418289185, 0.0026736625004559755, -0.026642367243766785, -0.06329245120286942, -0...
Spherical CNNs on Unstructured Grids
https://openreview.net/forum?id=Bkl-43C9FQ
[ "Chiyu Max Jiang", "Jingwei Huang", "Karthik Kashinath", "Prabhat", "Philip Marcus", "Matthias Niessner" ]
Poster
null
We present an efficient convolution kernel for Convolutional Neural Networks (CNNs) on unstructured grids using parameterized differential operators while focusing on spherical signals such as panorama images or planetary signals. To this end, we replace conventional convolution kernels with linear combinations of dif...
[ "Spherical CNN", "unstructured grid", "panoramic", "semantic segmentation", "parameter efficiency" ]
We present a new CNN kernel for unstructured grids for spherical signals, and show significant accuracy and parameter efficiency gain on tasks such as 3D classfication and omnidirectional image segmentation.
1,422
1901.02039
title_snapshot
[ 0.015382573008537292, -0.006000136025249958, 0.027634046971797943, 0.03567416965961456, 0.032240331172943115, 0.0363118015229702, 0.0010476479073986411, 0.007424451876431704, -0.030799871310591698, -0.05520537868142128, -0.014225345104932785, -0.016797125339508057, -0.0635475218296051, 0.0...
Boosting Robustness Certification of Neural Networks
https://openreview.net/forum?id=HJgeEh09KQ
[ "Gagandeep Singh", "Timon Gehr", "Markus Püschel", "Martin Vechev" ]
Poster
null
We present a novel approach for the certification of neural networks against adversarial perturbations which combines scalable overapproximation methods with precise (mixed integer) linear programming. This results in significantly better precision than state-of-the-art verifiers on challenging feedforward and convolut...
[ "Robustness certification", "Adversarial Attacks", "Abstract Interpretation", "MILP Solvers", "Verification of Neural Networks" ]
We refine the over-approximation results from incomplete verifiers using MILP solvers to prove more robustness properties than state-of-the-art.
1,421
null
null
[ -0.007589357905089855, -0.01601862721145153, -0.012795615941286087, 0.02935177832841873, 0.035422537475824356, 0.024004679173231125, 0.024885209277272224, -0.03217094764113426, -0.04588885232806206, -0.000982768600806594, 0.009846101514995098, 0.013583198189735413, -0.052020490169525146, 0...
Two-Timescale Networks for Nonlinear Value Function Approximation
https://openreview.net/forum?id=rJleN20qK7
[ "Wesley Chung", "Somjit Nath", "Ajin Joseph", "Martha White" ]
Poster
null
A key component for many reinforcement learning agents is to learn a value function, either for policy evaluation or control. Many of the algorithms for learning values, however, are designed for linear function approximation---with a fixed basis or fixed representation. Though there have been a few sound extensions to...
[ "Reinforcement learning", "policy evaluation", "nonlinear function approximation" ]
We propose an architecture for learning value functions which allows the use of any linear policy evaluation algorithm in tandem with nonlinear feature learning.
1,420
null
null
[ -0.02398870699107647, -0.011803671717643738, -0.0035431035794317722, 0.03230869770050049, 0.04357383772730827, 0.03514346852898598, 0.0075934226624667645, 0.00421157106757164, -0.03774033486843109, -0.022648686543107033, 0.025862272828817368, 0.004435902461409569, -0.06474947184324265, 0.0...
Differentiable Perturb-and-Parse: Semi-Supervised Parsing with a Structured Variational Autoencoder
https://openreview.net/forum?id=BJlgNh0qKQ
[ "Caio Corro", "Ivan Titov" ]
Poster
null
Human annotation for syntactic parsing is expensive, and large resources are available only for a fraction of languages. A question we ask is whether one can leverage abundant unlabeled texts to improve syntactic parsers, beyond just using the texts to obtain more generalisable lexical features (i.e. beyond word embed...
[ "differentiable dynamic programming", "variational auto-encoder", "dependency parsing", "semi-supervised learning" ]
Differentiable dynamic programming over perturbed input weights with application to semi-supervised VAE
1,419
1807.09875
title_snapshot
[ 0.0010966795962303877, -0.013765868730843067, -0.014209235087037086, 0.03423392400145531, 0.03349649906158447, 0.03271206468343735, 0.04010642319917679, 0.01038680411875248, -0.011459406465291977, -0.009093697182834148, -0.022044945508241653, 0.0149547029286623, -0.06857302784919739, 0.005...
Algorithmic Framework for Model-based Deep Reinforcement Learning with Theoretical Guarantees
https://openreview.net/forum?id=BJe1E2R5KX
[ "Yuping Luo", "Huazhe Xu", "Yuanzhi Li", "Yuandong Tian", "Trevor Darrell", "Tengyu Ma" ]
Poster
null
Model-based reinforcement learning (RL) is considered to be a promising approach to reduce the sample complexity that hinders model-free RL. However, the theoretical understanding of such methods has been rather limited. This paper introduces a novel algorithmic framework for designing and analyzing model-based RL algo...
[ "model-based reinforcement learning", "sample efficiency", "deep reinforcement learning" ]
We design model-based reinforcement learning algorithms with theoretical guarantees and achieve state-of-the-art results on Mujuco benchmark tasks when one million or fewer samples are permitted.
1,410
1807.03858
title_snapshot
[ -0.02693478763103485, -0.02021084725856781, 0.0028942706994712353, 0.02716328762471676, 0.036695875227451324, 0.013322453014552593, 0.01308666355907917, 0.006969413720071316, -0.039710741490125656, -0.012078925967216492, -0.022289087995886803, 0.007981423288583755, -0.07218505442142487, -0...
Tree-Structured Recurrent Switching Linear Dynamical Systems for Multi-Scale Modeling
https://openreview.net/forum?id=HkzRQhR9YX
[ "Josue Nassar", "Scott Linderman", "Monica Bugallo", "Il Memming Park" ]
Poster
null
Many real-world systems studied are governed by complex, nonlinear dynamics. By modeling these dynamics, we can gain insight into how these systems work, make predictions about how they will behave, and develop strategies for controlling them. While there are many methods for modeling nonlinear dynamical systems, exis...
[ "machine learning", "bayesian statistics", "dynamical systems" ]
null
1,408
1811.12386
title_snapshot
[ -0.02526894211769104, -0.029896993190050125, -0.0040768724866211414, 0.012301921844482422, 0.04340452700853348, 0.03481101989746094, 0.023284193128347397, -0.0196794755756855, -0.028435248881578445, -0.043659817427396774, 0.020219311118125916, -0.015100576914846897, -0.07065936177968979, -...
Diagnosing and Enhancing VAE Models
https://openreview.net/forum?id=B1e0X3C9tQ
[ "Bin Dai", "David Wipf" ]
Poster
null
Although variational autoencoders (VAEs) represent a widely influential deep generative model, many aspects of the underlying energy function remain poorly understood. In particular, it is commonly believed that Gaussian encoder/decoder assumptions reduce the effectiveness of VAEs in generating realistic samples. In ...
[ "variational autoencoder", "generative models" ]
We closely analyze the VAE objective function and draw novel conclusions that lead to simple enhancements.
1,405
1903.05789
title_snapshot
[ 0.022524187341332436, 0.014108502306044102, -0.011459838598966599, 0.052829667925834656, 0.008761685341596603, 0.047037169337272644, 0.05518452823162079, -0.0030299776699393988, -0.017851000651717186, -0.04600055515766144, -0.012506402097642422, 0.004412967246025801, -0.07051067799329758, ...
Efficiently testing local optimality and escaping saddles for ReLU networks
https://openreview.net/forum?id=HylTXn0qYX
[ "Chulhee Yun", "Suvrit Sra", "Ali Jadbabaie" ]
Poster
null
We provide a theoretical algorithm for checking local optimality and escaping saddles at nondifferentiable points of empirical risks of two-layer ReLU networks. Our algorithm receives any parameter value and returns: local minimum, second-order stationary point, or a strict descent direction. The presence of M data poi...
[ "local optimality", "second-order stationary point", "escaping saddle points", "nondifferentiability", "ReLU", "empirical risk" ]
A theoretical algorithm for testing local optimality and extracting descent directions at nondifferentiable points of empirical risks of one-hidden-layer ReLU networks.
1,404
1809.10858
title_snapshot
[ -0.04070805758237839, -0.02070002816617489, -0.009578459896147251, 0.06431709229946136, 0.03683886304497719, 0.054147057235240936, 0.0189107283949852, -0.0062175169587135315, -0.03256332129240036, -0.0412813238799572, 0.016638366505503654, 0.011818358674645424, -0.06248697265982628, -0.019...
Don't let your Discriminator be fooled
https://openreview.net/forum?id=HJE6X305Fm
[ "Brady Zhou", "Philipp Krähenbühl" ]
Poster
null
Generative Adversarial Networks are one of the leading tools in generative modeling, image editing and content creation. However, they are hard to train as they require a delicate balancing act between two deep networks fighting a never ending duel. Some of the most promising adversarial models today minimize a Wasser...
[ "GAN", "generative models", "computer vision" ]
A discriminator that is not easily fooled by adversarial example makes GAN training more robust and leads to a smoother objective.
1,403
null
null
[ -0.023435018956661224, -0.022761983796954155, -0.007062842138111591, 0.04732758551836014, 0.0030307467095553875, -0.0008169565699063241, 0.018431268632411957, 0.003935251850634813, 0.0007363809854723513, -0.052066996693611145, -0.03476683050394058, -0.009942349046468735, -0.05705147609114647...
Rigorous Agent Evaluation: An Adversarial Approach to Uncover Catastrophic Failures
https://openreview.net/forum?id=B1xhQhRcK7
[ "Jonathan Uesato*", "Ananya Kumar*", "Csaba Szepesvari*", "Tom Erez", "Avraham Ruderman", "Keith Anderson", "Krishnamurthy (Dj) Dvijotham", "Nicolas Heess", "Pushmeet Kohli" ]
Poster
null
This paper addresses the problem of evaluating learning systems in safety critical domains such as autonomous driving, where failures can have catastrophic consequences. We focus on two problems: searching for scenarios when learned agents fail and assessing their probability of failure. The standard method for agent e...
[ "agent evaluation", "adversarial examples", "robustness", "safety", "reinforcement learning" ]
We show that rare but catastrophic failures may be missed entirely by random testing, which poses issues for safe deployment. Our proposed approach for adversarial testing fixes this.
1,393
1812.01647
title_snapshot
[ -0.013591300696134567, -0.009072325192391872, -0.013257223181426525, 0.07074469327926636, 0.028859909623861313, 0.012909145094454288, 0.04052456468343735, -0.01072915829718113, -0.006685121916234493, -0.03156183287501335, -0.02679149992763996, 0.04382355511188507, -0.06370359659194946, -0....
Competitive experience replay
https://openreview.net/forum?id=Sklsm20ctX
[ "Hao Liu", "Alexander Trott", "Richard Socher", "Caiming Xiong" ]
Poster
null
Deep learning has achieved remarkable successes in solving challenging reinforcement learning (RL) problems when dense reward function is provided. However, in sparse reward environment it still often suffers from the need to carefully shape reward function to guide policy optimization. This limits the applicability of...
[ "reinforcement learning", "sparse reward", "goal-based learning" ]
a novel method to learn with sparse reward using adversarial reward re-labeling
1,391
1902.00528
title_snapshot
[ -0.029695264995098114, -0.020678794011473656, -0.009414834901690483, 0.02894400991499424, 0.0621405728161335, 0.024107782170176506, 0.017229294404387474, 0.01001558918505907, -0.05976414680480957, -0.030048474669456482, 0.011195880360901356, 0.022137587890028954, -0.029216744005680084, -0....
A Max-Affine Spline Perspective of Recurrent Neural Networks
https://openreview.net/forum?id=BJej72AqF7
[ "Zichao Wang", "Randall Balestriero", "Richard Baraniuk" ]
Poster
null
We develop a framework for understanding and improving recurrent neural networks (RNNs) using max-affine spline operators (MASOs). We prove that RNNs using piecewise affine and convex nonlinearities can be written as a simple piecewise affine spline operator. The resulting representation provides several new perspectiv...
[ "RNN", "max-affine spline operators" ]
We provide new insights and interpretations of RNNs from a max-affine spline operators perspective.
1,390
null
null
[ -0.0470767505466938, -0.007385254837572575, -0.002089882269501686, 0.04526389762759209, 0.032456789165735245, 0.06128450855612755, 0.044156596064567566, 0.017236117273569107, -0.05995495989918709, -0.042026713490486145, -0.005979697685688734, -0.006261669099330902, -0.05554421991109848, -0...
Top-Down Neural Model For Formulae
https://openreview.net/forum?id=Byg5QhR5FQ
[ "Karel Chvalovský" ]
Poster
null
We present a simple neural model that given a formula and a property tries to answer the question whether the formula has the given property, for example whether a propositional formula is always true. The structure of the formula is captured by a feedforward neural network recursively built for the given formula in a ...
[ "logic", "formula", "recursive neural networks", "recurrent neural networks" ]
A top-down approach how to recursively represent propositional formulae by neural networks is presented.
1,384
null
null
[ -0.03403312712907791, 0.04034895449876785, 0.006399382837116718, -0.0033253352157771587, 0.04795541241765022, 0.011856828816235065, -0.011459369212388992, 0.0002337716578040272, -0.02124430611729622, -0.009112372994422913, -0.022563813254237175, 0.025436557829380035, -0.05570872128009796, ...
Feature-Wise Bias Amplification
https://openreview.net/forum?id=S1ecm2C9K7
[ "Klas Leino", "Emily Black", "Matt Fredrikson", "Shayak Sen", "Anupam Datta" ]
Poster
null
We study the phenomenon of bias amplification in classifiers, wherein a machine learning model learns to predict classes with a greater disparity than the underlying ground truth. We demonstrate that bias amplification can arise via inductive bias in gradient descent methods resulting in overestimation of importance of...
[ "bias", "bias amplification", "classification" ]
null
1,382
1812.08999
title_snapshot
[ 0.022092586383223534, -0.03203324228525162, -0.009692342951893806, 0.009015556424856186, 0.03665073215961456, 0.03626026585698128, 0.05033749341964722, -0.015041535720229149, -0.02066083624958992, -0.04343895614147186, 0.016343412920832634, 0.018551098182797432, -0.09608953446149826, 0.023...
AD-VAT: An Asymmetric Dueling mechanism for learning Visual Active Tracking
https://openreview.net/forum?id=HkgYmhR9KX
[ "Fangwei Zhong", "Peng Sun", "Wenhan Luo", "Tingyun Yan", "Yizhou Wang" ]
Poster
null
Visual Active Tracking (VAT) aims at following a target object by autonomously controlling the motion system of a tracker given visual observations. Previous work has shown that the tracker can be trained in a simulator via reinforcement learning and deployed in real-world scenarios. However, during training, such a me...
[ "Active tracking", "reinforcement learning", "adversarial learning", "multi agent" ]
We propose AD-VAT, where the tracker and the target object, viewed as two learnable agents, are opponents and can mutually enhance during training.
1,378
null
null
[ 0.0148715665563941, -0.026764506474137306, 0.0025815609842538834, 0.03623109310865402, -0.013788001611828804, 0.008638984523713589, 0.010279574431478977, -0.022873280569911003, -0.0407002717256546, -0.05749749392271042, -0.03679027408361435, 0.007242836058139801, -0.07105743885040283, -0.0...
Learning to Learn with Conditional Class Dependencies
https://openreview.net/forum?id=BJfOXnActQ
[ "Xiang Jiang", "Mohammad Havaei", "Farshid Varno", "Gabriel Chartrand", "Nicolas Chapados", "Stan Matwin" ]
Poster
null
Neural networks can learn to extract statistical properties from data, but they seldom make use of structured information from the label space to help representation learning. Although some label structure can implicitly be obtained when training on huge amounts of data, in a few-shot learning context where little data...
[ "meta-learning", "learning to learn", "few-shot learning" ]
CAML is an instance of MAML with conditional class dependencies.
1,374
null
null
[ -0.013448444195091724, -0.0007724555325694382, -0.013055337592959404, 0.04581572860479355, 0.030865982174873352, 0.029745306819677353, 0.017049577087163925, -0.027041541412472725, -0.044750992208719254, 0.02271292917430401, -0.044742535799741745, 0.05097344145178795, -0.06594026833772659, ...
GAN Dissection: Visualizing and Understanding Generative Adversarial Networks
https://openreview.net/forum?id=Hyg_X2C5FX
[ "David Bau", "Jun-Yan Zhu", "Hendrik Strobelt", "Bolei Zhou", "Joshua B. Tenenbaum", "William T. Freeman", "Antonio Torralba" ]
Poster
null
Generative Adversarial Networks (GANs) have recently achieved impressive results for many real-world applications, and many GAN variants have emerged with improvements in sample quality and training stability. However, visualization and understanding of GANs is largely missing. How does a GAN represent our visual world...
[ "GANs", "representation", "interpretability", "causality" ]
GAN representations are examined in detail, and sets of representation units are found that control the generation of semantic concepts in the output.
1,371
1811.10597
title_snapshot
[ 0.008711992762982845, -0.009730776771903038, -0.033590927720069885, 0.04173427075147629, 0.0043554892763495445, 0.004609271883964539, 0.0002652056282386184, 0.010828133672475815, -0.009974171407520771, -0.05704903602600098, -0.035331301391124725, -0.00875025987625122, -0.07275331020355225, ...
TimbreTron: A WaveNet(CycleGAN(CQT(Audio))) Pipeline for Musical Timbre Transfer
https://openreview.net/forum?id=S1lvm305YQ
[ "Sicong Huang", "Qiyang Li", "Cem Anil", "Xuchan Bao", "Sageev Oore", "Roger B. Grosse" ]
Poster
null
In this work, we address the problem of musical timbre transfer, where the goal is to manipulate the timbre of a sound sample from one instrument to match another instrument while preserving other musical content, such as pitch, rhythm, and loudness. In principle, one could apply image-based style transfer techniques t...
[ "Generative models", "Timbre Transfer", "Wavenet", "CycleGAN" ]
We present the TimbreTron, a pipeline for perfoming high-quality timbre transfer on musical waveforms using CQT-domain style transfer.
1,370
1811.09620
title_snapshot
[ 0.007338236551731825, -0.03598790615797043, -0.031063180416822433, 0.018503578379750252, 0.04487274959683418, 0.004344242159277201, 0.006675095297396183, 0.00115205068141222, -0.017252953723073006, -0.06573878228664398, 0.0009208728442899883, 0.007151918485760689, -0.057136062532663345, 0....
A Mean Field Theory of Batch Normalization
https://openreview.net/forum?id=SyMDXnCcF7
[ "Greg Yang", "Jeffrey Pennington", "Vinay Rao", "Jascha Sohl-Dickstein", "Samuel S. Schoenholz" ]
Poster
null
We develop a mean field theory for batch normalization in fully-connected feedforward neural networks. In so doing, we provide a precise characterization of signal propagation and gradient backpropagation in wide batch-normalized networks at initialization. Our theory shows that gradient signals grow exponentially in d...
[ "theory", "batch normalization", "mean field theory", "trainability" ]
Batch normalization causes exploding gradients in vanilla feedforward networks.
1,368
1902.08129
title_snapshot
[ -0.03544478863477707, -0.012077191844582558, 0.00797174870967865, 0.02593948133289814, 0.04860749840736389, 0.03946574777364731, 0.023693053051829338, 0.019384510815143585, -0.03841247782111168, -0.045949455350637436, 0.010450270026922226, 0.007651954889297485, -0.03734979033470154, -0.008...
A Closer Look at Few-shot Classification
https://openreview.net/forum?id=HkxLXnAcFQ
[ "Wei-Yu Chen", "Yen-Cheng Liu", "Zsolt Kira", "Yu-Chiang Frank Wang", "Jia-Bin Huang" ]
Poster
null
Few-shot classification aims to learn a classifier to recognize unseen classes during training with limited labeled examples. While significant progress has been made, the growing complexity of network designs, meta-learning algorithms, and differences in implementation details make a fair comparison difficult. In this pap...
[ "few shot classification", "meta-learning" ]
A detailed empirical study in few-shot classification that revealing challenges in standard evaluation setting and showing a new direction.
1,361
1904.04232
title_snapshot
[ 0.018472684547305107, -0.05140501260757446, 0.002364960266277194, 0.051583632826805115, 0.02243977040052414, -0.0014364075614139438, 0.042970962822437286, -0.016926711425185204, -0.024695446714758873, -0.005323462653905153, -0.003919243812561035, 0.032353851944208145, -0.06496278196573257, ...
STCN: Stochastic Temporal Convolutional Networks
https://openreview.net/forum?id=HkzSQhCcK7
[ "Emre Aksan", "Otmar Hilliges" ]
Poster
null
Convolutional architectures have recently been shown to be competitive on many sequence modelling tasks when compared to the de-facto standard of recurrent neural networks (RNNs) while providing computational and modelling advantages due to inherent parallelism. However, currently, there remains a performance gap to mo...
[ "latent variables", "variational inference", "temporal convolutional networks", "sequence modeling", "auto-regressive modeling" ]
We combine the computational advantages of temporal convolutional architectures with the expressiveness of stochastic latent variables.
1,356
1902.06568
title_snapshot
[ 0.020168893039226532, -0.0150423189625144, -0.022112218663096428, 0.05071628466248512, 0.02180790901184082, 0.03902637958526611, 0.012042243964970112, 0.05702140927314758, -0.015618355944752693, -0.028567828238010406, 0.02382262609899044, -0.030240075662732124, -0.03562641516327858, 0.0177...
RotatE: Knowledge Graph Embedding by Relational Rotation in Complex Space
https://openreview.net/forum?id=HkgEQnRqYQ
[ "Zhiqing Sun", "Zhi-Hong Deng", "Jian-Yun Nie", "Jian Tang" ]
Poster
null
We study the problem of learning representations of entities and relations in knowledge graphs for predicting missing links. The success of such a task heavily relies on the ability of modeling and inferring the patterns of (or between) the relations. In this paper, we present a new approach for knowledge graph embeddi...
[ "knowledge graph embedding", "knowledge graph completion", "adversarial sampling" ]
A new state-of-the-art approach for knowledge graph embedding.
1,347
1902.10197
title_snapshot
[ 0.009685692377388477, -0.00656426977366209, 0.020861275494098663, 0.05136573314666748, 0.0421258769929409, -0.00019310294010210782, 0.03449945151805878, -0.032421644777059555, 0.011856349185109138, -0.012158047407865524, -0.027814343571662903, 0.023326078429818153, -0.06381752341985703, 0....
Learning to Navigate the Web
https://openreview.net/forum?id=BJemQ209FQ
[ "Izzeddin Gur", "Ulrich Rueckert", "Aleksandra Faust", "Dilek Hakkani-Tur" ]
Poster
null
Learning in environments with large state and action spaces, and sparse rewards, can hinder a Reinforcement Learning (RL) agent’s learning through trial-and-error. For instance, following natural language instructions on the Web (such as booking a flight ticket) leads to RL settings where input vocabulary and number of...
[ "navigating web pages", "reinforcement learning", "q learning", "curriculum learning", "meta training" ]
We train reinforcement learning policies using reward augmentation, curriculum learning, and meta-learning to successfully navigate web pages.
1,342
1812.09195
title_snapshot
[ -0.029591290280222893, -0.027155306190252304, -0.011711853556334972, 0.045451853424310684, 0.04387079179286957, -0.02464154362678528, 0.02274363674223423, 0.02100786566734314, -0.004128606058657169, -0.013285514898598194, -0.04935629293322563, 0.04621327668428421, -0.07109188288450241, -0....
Modeling Uncertainty with Hedged Instance Embeddings
https://openreview.net/forum?id=r1xQQhAqKX
[ "Seong Joon Oh", "Kevin P. Murphy", "Jiyan Pan", "Joseph Roth", "Florian Schroff", "Andrew C. Gallagher" ]
Poster
null
Instance embeddings are an efficient and versatile image representation that facilitates applications like recognition, verification, retrieval, and clustering. Many metric learning methods represent the input as a single point in the embedding space. Often the distance between points is used as a proxy for match confi...
[ "uncertainty", "instance embedding", "metric learning", "probabilistic embedding" ]
The paper proposes using probability distributions instead of points for instance embeddings tasks such as recognition and verification.
1,341
null
null
[ 0.015170463360846043, -0.01847652532160282, -0.02560241147875786, 0.07058919966220856, 0.03606271371245384, 0.04036996141076088, 0.009361319243907928, -0.0035564147401601076, -0.011098898015916348, -0.052550096064805984, -0.022424768656492233, 0.004517205525189638, -0.07970693707466125, 0....
Automatically Composing Representation Transformations as a Means for Generalization
https://openreview.net/forum?id=B1ffQnRcKX
[ "Michael Chang", "Abhishek Gupta", "Sergey Levine", "Thomas L. Griffiths" ]
Poster
null
A generally intelligent learner should generalize to more complex tasks than it has previously encountered, but the two common paradigms in machine learning -- either training a separate learner per task or training a single learner for all tasks -- both have difficulty with such generalization because they do not leve...
[ "compositionality", "deep learning", "metareasoning" ]
We explore the problem of compositional generalization and propose a means for endowing neural network architectures with the ability to compose themselves to solve these problems.
1,340
1807.04640
title_snapshot
[ -0.0007424294017255306, -0.008443349972367287, -0.028727378696203232, 0.04204301908612251, 0.0397079735994339, 0.020690543577075005, 0.021504521369934082, 0.004565100651234388, -0.018530014902353287, -0.01375412754714489, -0.03656098246574402, 0.012747972272336483, -0.0734785720705986, 0.0...
Systematic Generalization: What Is Required and Can It Be Learned?
https://openreview.net/forum?id=HkezXnA9YX
[ "Dzmitry Bahdanau*", "Shikhar Murty*", "Michael Noukhovitch", "Thien Huu Nguyen", "Harm de Vries", "Aaron Courville" ]
Poster
null
Numerous models for grounded language understanding have been recently proposed, including (i) generic models that can be easily adapted to any given task and (ii) intuitively appealing modular models that require background knowledge to be instantiated. We compare both types of models in how much they lend themselves ...
[ "systematic generalization", "language understanding", "visual questions answering", "neural module networks" ]
We show that modular structured models are the best in terms of systematic generalization and that their end-to-end versions don't generalize as well.
1,335
1811.12889
title_snapshot
[ -0.0014360062777996063, -0.007318662945181131, -0.015181662514805794, 0.03001408651471138, 0.04319218173623085, 0.014205084182322025, 0.059608858078718185, -0.005761783570051193, -0.039058998227119446, -0.007307595107704401, -0.03751436620950699, 0.06847180426120758, -0.06562590599060059, ...
Adaptive Input Representations for Neural Language Modeling
https://openreview.net/forum?id=ByxZX20qFQ
[ "Alexei Baevski", "Michael Auli" ]
Poster
null
We introduce adaptive input representations for neural language modeling which extend the adaptive softmax of Grave et al. (2017) to input representations of variable capacity. There are several choices on how to factorize the input and output layers, and whether to model words, characters or sub-word units. We perform...
[ "Neural language modeling" ]
Variable capacity input word embeddings and SOTA on WikiText-103, Billion Word benchmarks.
1,334
1809.10853
title_snapshot
[ -0.030083518475294113, -0.023141538724303246, 0.004480919800698757, 0.024184666574001312, 0.022763866931200027, 0.07577338069677353, 0.019187062978744507, 0.016710281372070312, -0.015380769036710262, -0.012384019792079926, -0.024387413635849953, 0.012449787929654121, -0.040080051869153976, ...
Optimal Control Via Neural Networks: A Convex Approach
https://openreview.net/forum?id=H1MW72AcK7
[ "Yize Chen", "Yuanyuan Shi", "Baosen Zhang" ]
Poster
null
Control of complex systems involves both system identification and controller design. Deep neural networks have proven to be successful in many identification tasks, however, from model-based control perspective, these networks are difficult to work with because they are typically nonlinear and nonconvex. Therefore man...
[ "optimal control", "input convex neural network", "convex optimization" ]
null
1,332
1805.11835
title_snapshot
[ -0.047523967921733856, 0.002905180910602212, -0.02532760426402092, 0.036754753440618515, 0.06250268965959549, 0.027828199788928032, 0.011433238163590431, 0.00998544692993164, -0.02884158305823803, -0.052451275289058685, 0.000657885626424104, -0.018447743728756905, -0.04575667530298233, -0....
An Empirical Study of Example Forgetting during Deep Neural Network Learning
https://openreview.net/forum?id=BJlxm30cKm
[ "Mariya Toneva*", "Alessandro Sordoni*", "Remi Tachet des Combes*", "Adam Trischler", "Yoshua Bengio", "Geoffrey J. Gordon" ]
Poster
null
Inspired by the phenomenon of catastrophic forgetting, we investigate the learning dynamics of neural networks as they train on single classification tasks. Our goal is to understand whether a related phenomenon occurs when data does not undergo a clear distributional shift. We define a ``forgetting event'' to have occ...
[ "catastrophic forgetting", "sample weighting", "deep generalization" ]
We show that catastrophic forgetting occurs within what is considered to be a single task and find that examples that are not prone to forgetting can be removed from the training set without loss of generalization.
1,326
1812.05159
title_snapshot
[ -0.04020778834819794, 0.00026441121008247137, -0.02557561732828617, 0.041121795773506165, 0.04256102815270424, 0.006814824417233467, 0.014176642522215843, 0.027599966153502464, -0.058111727237701416, -0.015292261727154255, -0.004240322392433882, 0.01721273735165596, -0.04765750840306282, 0...
GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding
https://openreview.net/forum?id=rJ4km2R5t7
[ "Alex Wang", "Amanpreet Singh", "Julian Michael", "Felix Hill", "Omer Levy", "Samuel R. Bowman" ]
Poster
null
For natural language understanding (NLU) technology to be maximally useful, it must be able to process language in a way that is not exclusive to a single task, genre, or dataset. In pursuit of this objective, we introduce the General Language Understanding Evaluation (GLUE) benchmark, a collection of tools for evaluat...
[ "natural language understanding", "multi-task learning", "evaluation" ]
We present a multi-task benchmark and analysis platform for evaluating generalization in natural language understanding systems.
1,323
1804.07461
title_snapshot
[ -0.005377187859266996, 0.0064489278011024, -0.02892567217350006, 0.036587607115507126, 0.03444647416472435, 0.04921197146177292, 0.03162792697548866, 0.017213741317391396, -0.0026522637344896793, 0.00007662667485419661, -0.029626980423927307, 0.05684925988316536, -0.07547220587730408, -0.0...
Plan Online, Learn Offline: Efficient Learning and Exploration via Model-Based Control
https://openreview.net/forum?id=Byey7n05FQ
[ "Kendall Lowrey", "Aravind Rajeswaran", "Sham Kakade", "Emanuel Todorov", "Igor Mordatch" ]
Poster
null
We propose a "plan online and learn offline" framework for the setting where an agent, with an internal model, needs to continually act and learn in the world. Our work builds on the synergistic relationship between local model-based control, global value function learning, and exploration. We study how local trajector...
[ "deep reinforcement learning", "exploration", "model-based" ]
We propose a framework that incorporates planning for efficient exploration and learning in complex environments.
1,318
1811.01848
title_snapshot
[ -0.025568852201104164, -0.0038910098373889923, -0.014786791987717152, 0.04167424142360687, 0.06076781451702118, 0.017445562407374382, 0.004670644644647837, -0.0046708667650818825, -0.028704781085252762, -0.030268890783190727, -0.02651526778936386, 0.013245477341115475, -0.05683363229036331, ...
Learning Grid Cells as Vector Representation of Self-Position Coupled with Matrix Representation of Self-Motion
https://openreview.net/forum?id=Syx0Mh05YQ
[ "Ruiqi Gao", "Jianwen Xie", "Song-Chun Zhu", "Ying Nian Wu" ]
Poster
null
This paper proposes a representational model for grid cells. In this model, the 2D self-position of the agent is represented by a high-dimensional vector, and the 2D self-motion or displacement of the agent is represented by a matrix that transforms the vector. Each component of the vector is a unit or a cell. The mode...
[]
null
1,317
1810.05597
title_snapshot
[ -0.026261072605848312, -0.009772997349500656, -0.007226424291729927, 0.0227137990295887, 0.03527332469820976, 0.014173317700624466, -0.0019251066260039806, -0.0015105541097000241, -0.04787022992968559, -0.04805630445480347, -0.0028948187828063965, -0.04565761238336563, -0.05708111450076103, ...
Preventing Posterior Collapse with delta-VAEs
https://openreview.net/forum?id=BJe0Gn0cY7
[ "Ali Razavi", "Aaron van den Oord", "Ben Poole", "Oriol Vinyals" ]
Poster
null
Due to the phenomenon of “posterior collapse,” current latent variable generative models pose a challenging design choice that either weakens the capacity of the decoder or requires altering the training objective. We develop an alternative that utilizes the most powerful generative models as decoders, optimize the var...
[ "Posterior Collapse", "VAE", "Autoregressive Models" ]
Avoid posterior collapse by lower bounding the rate.
1,316
1901.03416
title_snapshot
[ 0.043035589158535004, 0.02566201612353325, -0.02243042178452015, 0.046928539872169495, 0.049052219837903976, 0.06551489979028702, 0.04423965513706207, -0.024173934012651443, -0.023572387173771858, -0.047340601682662964, -0.013200776651501656, 0.0018219816265627742, -0.05692723020911217, 0....
Deep Online Learning Via Meta-Learning: Continual Adaptation for Model-Based RL
https://openreview.net/forum?id=HyxAfnA5tm
[ "Anusha Nagabandi", "Chelsea Finn", "Sergey Levine" ]
Poster
null
Humans and animals can learn complex predictive models that allow them to accurately and reliably reason about real-world phenomena, and they can adapt such models extremely quickly in the face of unexpected changes. Deep neural network models allow us to represent very complex functions, but lack this capacity for rap...
[ "meta-learning", "model-based", "reinforcement learning", "online learning", "adaptation" ]
null
1,315
1812.07671
title_snapshot
[ -0.019346123561263084, -0.000623162486590445, -0.029249679297208786, 0.021961254999041557, 0.05370596423745155, 0.03550741821527481, 0.02484147623181343, 0.03552127256989479, -0.04544631391763687, -0.006915729027241468, -0.021147817373275757, 0.025631790980696678, -0.07332304120063782, -0....
Value Propagation Networks
https://openreview.net/forum?id=SJG6G2RqtX
[ "Nantas Nardelli", "Gabriel Synnaeve", "Zeming Lin", "Pushmeet Kohli", "Philip H. S. Torr", "Nicolas Usunier" ]
Poster
null
We present Value Propagation (VProp), a set of parameter-efficient differentiable planning modules built on Value Iteration which can successfully be trained using reinforcement learning to solve unseen tasks, has the capability to generalize to larger map sizes, and can learn to navigate in dynamic environments. We sh...
[ "Reinforcement Learning", "Value Iteration", "Navigation", "Convolutional Neural Networks", "Learning to plan" ]
We present planners based on convnets that are sample-efficient and that generalize to larger instances of navigation and pathfinding problems.
1,308
1805.11199
title_snapshot
[ -0.012567892670631409, 0.009417752735316753, 0.018227925524115562, 0.04839722067117691, 0.048941899091005325, 0.03210505470633507, 0.008936951868236065, 0.0001318626309512183, -0.0415455661714077, -0.059615086764097214, -0.025968320667743683, 0.01237897202372551, -0.06942550092935562, -0.0...