title
stringlengths
16
162
paper_url
stringlengths
41
43
authors
listlengths
1
27
type
stringclasses
0 values
primary_area
stringclasses
0 values
abstract
large_stringlengths
393
2.47k
keywords
listlengths
0
13
TL;DR
large_stringlengths
4
250
submission_number
int64
14
13.1k
arxiv_id
stringlengths
10
10
arxiv_id_source
stringclasses
2 values
embedding
listlengths
768
768
Statistical Learning and Inverse Problems: A Stochastic Gradient Approach
https://openreview.net/forum?id=09QFnDWPF8
[ "Yuri Fonseca", "Yuri Saporito" ]
null
null
Inverse problems are paramount in Science and Engineering. In this paper, we consider the setup of Statistical Inverse Problem (SIP) and demonstrate how Stochastic Gradient Descent (SGD) algorithms can be used to solve linear SIP. We provide consistency and finite sample bounds for the excess risk. We also propose a mo...
[ "Statistical Learning", "Inverse Problems", "Stochastic Gradient Descent" ]
An algorithm based on stochastic gradient descent for solving linear Inverse Problems under a statistical learning framework.
13,051
2209.14967
title_snapshot
[ -0.01468218956142664, -0.02945557050406933, -0.01884179748594761, 0.05087677389383316, 0.06435997039079666, 0.0309649296104908, 0.032797861844301224, 0.006160533521324396, -0.025319518521428108, -0.029345888644456863, -0.0067244707606732845, -0.019756238907575607, -0.04864368960261345, -0....
Efficiency Ordering of Stochastic Gradient Descent
https://openreview.net/forum?id=pnSyqRXx73
[ "Jie Hu", "Vishwaraj Doshi", "Do Young Eun" ]
null
null
We consider the stochastic gradient descent (SGD) algorithm driven by a general stochastic sequence, including i.i.d noise and random walk on an arbitrary graph, among others; and analyze it in the asymptotic sense. Specifically, we employ the notion of `efficiency ordering', a well-analyzed tool for comparing the perf...
[ "Stochastic Gradient Descent", "Asymptotic Analysis", "Efficiency Ordering" ]
We introduce the notion of efficiency ordering as an alternative metric for comparing the performance of different stochastic input sequences for Stochastic Gradient Descent algorithm.
13,028
2209.07446
title_snapshot
[ -0.056077007204294205, -0.013329139910638332, -0.008249666541814804, 0.056267447769641876, 0.022250132635235786, 0.030613087117671967, 0.036126524209976196, 0.0032136542722582817, -0.005324619356542826, -0.04603495076298714, 0.01782214641571045, -0.008271371945738792, -0.05144507437944412, ...
Self-Aware Personalized Federated Learning
https://openreview.net/forum?id=EqJ5_hZSqgy
[ "Huili Chen", "Jie Ding", "Eric William Tramel", "Shuang Wu", "Anit Kumar Sahu", "Salman Avestimehr", "Tao Zhang" ]
null
null
In the context of personalized federated learning (FL), the critical challenge is to balance local model improvement and global model tuning when the personal and global objectives may not be exactly aligned. Inspired by Bayesian hierarchical models, we develop a self-aware personalized FL method where each client can ...
[ "Federared Learning", "Personalization" ]
We propose a new adaptive federated learning algorithm for personalization
13,014
2204.08069
title_snapshot
[ 0.016953200101852417, -0.05798129364848137, 0.02216949686408043, 0.015020253136754036, 0.054220087826251984, 0.05325939506292343, 0.043143268674612045, -0.013863502070307732, -0.012389184907078743, -0.04893064126372337, 0.008400321006774902, -0.004959167446941137, -0.04228920862078667, -0....
Nonnegative Tensor Completion via Integer Optimization
https://openreview.net/forum?id=xnI37HyfoP
[ "Caleb Xavier Bugg", "Chen Chen", "Anil Aswani" ]
null
null
Unlike matrix completion, tensor completion does not have an algorithm that is known to achieve the information-theoretic sample complexity rate. This paper develops a new algorithm for the special case of completion for nonnegative tensors. We prove that our algorithm converges in a linear (in numerical tolerance) num...
[ "tensor completion", "machine learning" ]
We present a new norm for nonnegative tensor completion and demonstrate its usefulness, versus existing methods, through numerical experiments
13,003
2111.04580
title_snapshot
[ -0.02381785586476326, -0.013528057374060154, 0.028729364275932312, 0.03735406696796417, 0.009257680736482143, 0.024362696334719658, 0.0029744612984359264, -0.013863218016922474, -0.03801402822136879, -0.05276622623205185, -0.03786107525229454, 0.0165152158588171, -0.05495305359363556, 0.00...
TPU-KNN: K Nearest Neighbor Search at Peak FLOP/s
https://openreview.net/forum?id=OoNmOfYVhEU
[ "Felix Chern", "Blake Hechtman", "Andy Davis", "Ruiqi Guo", "David Majnemer", "Sanjiv Kumar" ]
null
null
This paper presents a novel nearest neighbor search algorithm achieving TPU (Google Tensor Processing Unit) peak performance, outperforming state-of-the-art GPU algorithms with similar level of recall. The design of the proposed algorithm is motivated by an accurate accelerator performance model that takes into account...
[ "TPU", "K-nearest neighbor search", "Approximate nearest neighbor search", "roofline model", "accelerator" ]
Novel nearest neighbor search algorithm achieving TPU peak performance with recall guarantee.
12,999
2206.14286
title_snapshot
[ -0.03889736160635948, -0.04195167124271393, -0.010710516013205051, 0.04164697229862213, 0.012290875427424908, 0.041699182242155075, 0.01304577011615038, 0.02166103757917881, -0.012340549379587173, -0.04015690088272095, 0.0068510184064507484, -0.03146093711256981, -0.05832378566265106, 0.03...
Equivariant Networks for Crystal Structures
https://openreview.net/forum?id=0Dh8dz4snu
[ "Sékou-Oumar Kaba", "Siamak Ravanbakhsh" ]
null
null
Supervised learning with deep models has tremendous potential for applications in materials science. Recently, graph neural networks have been used in this context, drawing direct inspiration from models for molecules. However, materials are typically much more structured than molecules, which is a feature that these m...
[ "materials", "deep learning", "symmetry", "equivariance", "crystals", "graph neural networks", "geometric deep learning" ]
A deep model for materials
12,977
2211.15420
title_snapshot
[ 0.03147461265325546, 0.013206856325268745, -0.018281491473317146, 0.04047037288546562, 0.022125815972685814, -0.021732987836003304, 0.02495388314127922, 0.013800445944070816, -0.009212865494191647, -0.04556659981608391, 0.0011669463710859418, -0.017837652936577797, -0.04916859418153763, 0....
Gradient Descent Is Optimal Under Lower Restricted Secant Inequality And Upper Error Bound
https://openreview.net/forum?id=s1yaWFDLxVG
[ "Charles Guille-Escuret", "Adam Ibrahim", "Baptiste Goujaud", "Ioannis Mitliagkas" ]
null
null
The study of first-order optimization is sensitive to the assumptions made on the objective functions. These assumptions induce complexity classes which play a key role in worst-case analysis, including the fundamental concept of algorithm optimality. Recent work argues that strong convexity and smoothness—popular assu...
[ "First-Order Optimization", "Non-Convex", "Deterministic", "Gradient Descent", "Restricted Secant Inequality", "Error Bounds" ]
We show that Gradient Descent is exactly optimal on a class of functions relevant to machine learning using Performance Estimation Problems
12,967
2203.00342
title_snapshot
[ -0.058542851358652115, 0.014764049090445042, 0.026872804388403893, 0.029763055965304375, 0.05129345506429672, 0.031176719814538956, 0.03830798342823982, -0.02645430713891983, -0.022916294634342194, -0.02653813362121582, -0.025819867849349976, 0.0014294009888544679, -0.05670185014605522, 0....
Decoupled Context Processing for Context Augmented Language Modeling
https://openreview.net/forum?id=02dbnEbEFn
[ "Zonglin Li", "Ruiqi Guo", "Sanjiv Kumar" ]
null
null
Language models can be augmented with context retriever to incorporate knowledge from large external databases. By leveraging retrieved context, the neural network does not have to memorize the massive amount of world knowledge within its internal parameters, leading to better parameter efficiency, interpretability and...
[ "Retrieval Augmentation", "Encoder-Decoder", "Language Modeling", "Efficiency" ]
null
12,953
2210.05758
title_snapshot
[ 0.00138951709959656, 0.027684859931468964, -0.014839520677924156, 0.04280950501561165, 0.03474942594766617, 0.03744766488671303, 0.02672753296792507, 0.022971943020820618, -0.009188426658511162, -0.007374530658125877, -0.03838573023676872, 0.04635104164481163, -0.050776273012161255, -0.007...
Planning to the Information Horizon of BAMDPs via Epistemic State Abstraction
https://openreview.net/forum?id=7eUOC9fEIRO
[ "Dilip Arumugam", "Satinder Singh" ]
null
null
The Bayes-Adaptive Markov Decision Process (BAMDP) formalism pursues the Bayes-optimal solution to the exploration-exploitation trade-off in reinforcement learning. As the computation of exact solutions to Bayesian reinforcement-learning problems is intractable, much of the literature has focused on developing suitable...
[ "Bayes-Adaptive Markov Decision Process", "Bayesian reinforcement learning", "Exploration", "Planning" ]
null
12,944
2210.16872
title_snapshot
[ -0.05750054493546486, 0.010206039994955063, -0.027864331379532814, 0.02455195039510727, 0.0550476610660553, 0.016424285247921944, 0.02656242623925209, -0.026233602315187454, -0.03798150271177292, -0.024753758683800697, -0.023721890524029732, 0.021652715280652046, -0.04986030235886574, -0.0...
Trust Region Policy Optimization with Optimal Transport Discrepancies: Duality and Algorithm for Continuous Actions
https://openreview.net/forum?id=BUMiizPcby6
[ "Antonio Terpin", "Nicolas Lanzetti", "Batuhan Yardim", "Florian Dorfler", "Giorgia Ramponi" ]
null
null
Policy Optimization (PO) algorithms have been proven particularly suited to handle the high-dimensionality of real-world continuous control tasks. In this context, Trust Region Policy Optimization methods represent a popular approach to stabilize the policy updates. These usually rely on the Kullback-Leibler (KL) diver...
[ "Trust region policy optimization", "optimal transport" ]
null
12,920
2210.11137
title_snapshot
[ -0.04735783115029335, -0.00995560735464096, 0.012272664345800877, 0.08055464178323746, 0.0540829561650753, 0.010017747059464455, 0.03127121925354004, -0.009290354326367378, 0.0056961835362017155, -0.07035841792821884, -0.0066840858198702335, -0.002148533007130027, -0.06585676968097687, -0....
Modeling Transitivity and Cyclicity in Directed Graphs via Binary Code Box Embeddings
https://openreview.net/forum?id=kpSAfnHSgXR
[ "Dongxu Zhang", "Michael Boratko", "Cameron N Musco", "Andrew McCallum" ]
null
null
Modeling directed graphs with differentiable representations is a fundamental requirement for performing machine learning on graph-structured data. Geometric embedding models (e.g. hyperbolic, cone, and box embeddings) excel at this task, exhibiting useful inductive biases for directed graphs. However, modeling directe...
[ "graph representation learning", "geometric representation learning", "directed graphs", "cyclic graphs", "transitivity" ]
null
12,916
null
null
[ 0.017229299992322922, -0.019132351502776146, -0.018560266122221947, 0.0483362078666687, 0.006305872928351164, 0.010209056548774242, 0.04818079248070717, 0.04113295301795006, -0.004917966201901436, -0.05057761073112488, 0.009329128079116344, -0.03622984141111374, -0.07883182168006897, 0.011...
Simple and Optimal Greedy Online Contention Resolution Schemes
https://openreview.net/forum?id=qx51yfvLnE
[ "Vasilis Livanos" ]
null
null
Matching based markets, like ad auctions, ride-sharing, and eBay, are inherently online and combinatorial, and therefore have been extensively studied under the lens of online stochastic combinatorial optimization models. The general framework that has emerged uses Contention Resolution Schemes (CRSs) introduced by Che...
[ "contention resolution schemes", "online algorithms", "matroids", "prophet inequalities" ]
We give the first optimal greedy OCRS for the single-item setting, improving both upper and lower bounds to 1/e.
12,915
2111.13253
title_snapshot
[ -0.04152325168251991, -0.012815193273127079, 0.002625387627631426, 0.05304766446352005, 0.023204797878861427, 0.03648141026496887, -0.01190724316984415, 0.005661555100232363, -0.022756360471248627, -0.02677506022155285, -0.0036915128584951162, -0.007099078502506018, -0.07563687115907669, -...
Evaluating Latent Space Robustness and Uncertainty of EEG-ML Models under Realistic Distribution Shifts
https://openreview.net/forum?id=KRk0lBRPpOC
[ "Neeraj Wagh", "Jionghao Wei", "Samarth Rawal", "Brent M. Berry", "Yogatheesan Varatharajah" ]
null
null
The recent availability of large datasets in bio-medicine has inspired the development of representation learning methods for multiple healthcare applications. Despite advances in predictive performance, the clinical utility of such methods is limited when exposed to real-world data. This study develops model diagnosti...
[ "dataset shifts", "scalp EEG", "representation learning", "healthcare machine learning", "model robustness", "latent space", "uncertainty quantification", "distribution shift" ]
We develop model diagnostic measures to identify failure modes of EEG-ML models before deployment without access to out-of-distribution data.
12,911
2209.11233
title_snapshot
[ -0.015477445907890797, -0.011631024070084095, -0.03249824047088623, -0.0007056297617964447, 0.07037307322025299, 0.006500064395368099, 0.039741795510053635, -0.011013136245310307, -0.01877642795443535, -0.049434151500463486, -0.015653438866138458, -0.0032841218635439873, -0.06228864938020706...
COLD Decoding: Energy-based Constrained Text Generation with Langevin Dynamics
https://openreview.net/forum?id=TiZYrQ-mPup
[ "Lianhui Qin", "Sean Welleck", "Daniel Khashabi", "Yejin Choi" ]
null
null
Many applications of text generation require incorporating different constraints to control the semantics or style of generated text. These constraints can be hard (e.g., ensuring certain keywords are included in the output) and soft (e.g., contextualizing the output with the left- or right-hand context). In this paper...
[ "Text generation", "constrained text generation", "language models", "natural language processing", "langevin dynamics", "decoding" ]
null
12,908
2202.11705
title_snapshot
[ 0.0019162674434483051, 0.003881481708958745, -0.006165661849081516, 0.06162088364362717, 0.0709235668182373, 0.0340043380856514, 0.03522109612822533, 0.02313627116382122, -0.04214182123541832, -0.023982610553503036, -0.009842409752309322, 0.035074569284915924, -0.0810157060623169, -0.01871...
From Gradient Flow on Population Loss to Learning with Stochastic Gradient Descent
https://openreview.net/forum?id=xuw7R0hP7G
[ "Christopher De Sa", "Satyen Kale", "Jason D. Lee", "Ayush Sekhari", "Karthik Sridharan" ]
null
null
Stochastic Gradient Descent (SGD) has been the method of choice for learning large-scale non-convex models. While a general analysis of when SGD works has been elusive, there has been a lot of recent progress in understanding the convergence of Gradient Flow (GF) on the population loss, partly due to the simplicity th...
[ "Theory", "Gradient flow", "Stochastic Gradient Descent", "Gradient Descent", "SGD", "Non-convex optimization", "Lyapunov potentials" ]
We show that under mild assumptions, whenever gradient flow works on the population loss, stochastic gradient descent succeeds at learning.
12,907
2210.06705
title_snapshot
[ -0.046985380351543427, -0.03331359103322029, 0.013468514196574688, 0.041730113327503204, 0.0236335638910532, 0.04222126677632332, 0.018323803320527077, 0.02240889146924019, -0.025313494727015495, -0.03636107221245766, -0.002963458886370063, -0.02079703100025654, -0.06811782717704773, -0.00...
Fast Neural Kernel Embeddings for General Activations
https://openreview.net/forum?id=yLilJ1vZgMe
[ "Insu Han", "Amir Zandieh", "Jaehoon Lee", "Roman Novak", "Lechao Xiao", "Amin Karbasi" ]
null
null
Infinite width limit has shed light on generalization and optimization aspects of deep learning by establishing connections between neural networks and kernel methods. Despite their importance, the utility of these kernel methods was limited in large-scale learning settings due to their (super-)quadratic runtime and me...
[]
null
12,892
2209.04121
title_snapshot
[ -0.01825244165956974, -0.03782108426094055, 0.02801452949643135, 0.0318884514272213, 0.0029372419230639935, 0.05157046765089035, 0.016146132722496986, -0.006370927207171917, -0.010308406315743923, -0.03027361072599888, -0.009379010647535324, -0.004632212221622467, -0.0647374764084816, 0.00...
On Reinforcement Learning and Distribution Matching for Fine-Tuning Language Models with no Catastrophic Forgetting
https://openreview.net/forum?id=XvI6h-s4un
[ "Tomasz Korbak", "Hady Elsahar", "Germán Kruszewski", "Marc Dymetman" ]
null
null
The availability of large pre-trained models is changing the landscape of Machine Learning research and practice, moving from a "training from scratch" to a "fine-tuning'' paradigm. While in some applications the goal is to "nudge'' the pre-trained distribution towards preferred outputs, in others it is to steer it tow...
[ "Reinforcement Learning", "Language Models", "Reward Maximization", "Distribution Matching", "Energy Based Models", "Controlled Text Generation" ]
We describe and exploit connections between two distinct paradigms for expressing preferences over outputs of language models: reward maximization and distribution matching.
12,891
2206.00761
title_snapshot
[ -0.04817533493041992, -0.008540499955415726, -0.007433688268065453, 0.05127310752868652, 0.059793341904878616, 0.052986785769462585, 0.017970148473978043, -0.0024222121573984623, -0.03156734257936478, -0.006874526385217905, -0.03281761333346367, 0.0446041002869606, -0.06990832835435867, -0...
Provably tuning the ElasticNet across instances
https://openreview.net/forum?id=ZMFQtvVJr40
[ "Nina Balcan", "Mikhail Khodak", "Dravyansh Sharma", "Ameet Talwalkar" ]
null
null
An important unresolved challenge in the theory of regularization is to set the regularization coefficients of popular techniques like the ElasticNet with general provable guarantees. We consider the problem of tuning the regularization parameters of Ridge regression, LASSO, and the ElasticNet across multiple problem i...
[ "Elastic net", "data-driven algorithm design", "learning theory", "regularization" ]
We uncover structural properties of the ElasticNet that allow us to provably tune parameters given multiple problem instances - both online and in the statistical setting.
12,870
2207.10199
title_snapshot
[ -0.028392300009727478, -0.03801435977220535, 0.026149587705731392, 0.02292182296514511, 0.0705106258392334, 0.03266212344169617, 0.030115604400634766, -0.02304663509130478, -0.03843645751476288, -0.047198910266160965, -0.005336769390851259, 0.012780122458934784, -0.057877179235219955, -0.0...
LAMP: Extracting Text from Gradients with Language Model Priors
https://openreview.net/forum?id=6iqd9JAVR1z
[ "Mislav Balunovic", "Dimitar Iliev Dimitrov", "Nikola Jovanović", "Martin Vechev" ]
null
null
Recent work shows that sensitive user data can be reconstructed from gradient updates, breaking the key privacy promise of federated learning. While success was demonstrated primarily on image data, these methods do not directly transfer to other domains such as text. In this work, we propose LAMP, a novel attack tailo...
[ "federated learning", "privacy", "gradient leakage", "natural language processing" ]
We propose a novel attack for text reconstruction from gradients in federated learning based on language model priors.
12,868
2202.08827
title_snapshot
[ -0.0014960665721446276, -0.020736200734972954, -0.023453939706087112, 0.07347651571035385, 0.03903510794043541, 0.005155251361429691, 0.03345998376607895, 0.002111010951921344, -0.019213460385799408, -0.0073129781521856785, -0.018371516838669777, 0.017193375155329704, -0.0603477917611599, ...
ELIGN: Expectation Alignment as a Multi-Agent Intrinsic Reward
https://openreview.net/forum?id=uPyNR2yPoe
[ "Zixian Ma", "Rose E Wang", "Li Fei-Fei", "Michael S. Bernstein", "Ranjay Krishna" ]
null
null
Modern multi-agent reinforcement learning frameworks rely on centralized training and reward shaping to perform well. However, centralized training and dense rewards are not readily available in the real world. Current multi-agent algorithms struggle to learn in the alternative setup of decentralized training or sparse...
[ "Multi-agent intrinsic reward", "multi-agent reinforcement learning" ]
Inspired by the self-organization principle in Zoology, we introduce alignment, a task-agnostic and self-supervised intrinsic reward that encourages aligning dynamics: individual agents learn behaviors that match their neighbors' expectations.
12,865
2210.04365
title_snapshot
[ -0.025560790672898293, -0.014554698020219803, -0.01682005077600479, 0.053876638412475586, 0.01145733892917633, -0.000012743937986670062, 0.02830243855714798, -0.0020069952588528395, -0.03256719931960106, -0.04167342931032181, -0.03203466162085533, 0.013618596829473972, -0.07340972125530243, ...
Explicable Policy Search
https://openreview.net/forum?id=82N_rasrUT_
[ "Ze Gong", "Yu Zhang" ]
null
null
Human teammates often form conscious and subconscious expectations of each other during interaction. Teaming success is contingent on whether such expectations can be met. Similarly, for an intelligent agent to operate beside a human, it must consider the human’s expectation of its behavior. Disregarding such expectati...
[ "Explainable Decision Making", "Human-Aware AI" ]
null
12,862
null
null
[ -0.03631893917918205, 0.0068295905366539955, -0.030629023909568787, 0.04767555743455887, 0.049361441284418106, -0.026831194758415222, 0.018615031614899635, 0.000008184764737961814, -0.014585326425731182, -0.051055025309324265, -0.0314239040017128, 0.019403673708438873, -0.07485974580049515, ...
A Practical, Progressively-Expressive GNN
https://openreview.net/forum?id=WBv9Z6qpA8x
[ "Lingxiao Zhao", "Neil Shah", "Leman Akoglu" ]
null
null
Message passing neural networks (MPNNs) have become a dominant flavor of graph neural networks (GNNs) in recent years. Yet, MPNNs come with notable limitations; namely, they are at most as powerful as the 1-dimensional Weisfeiler-Leman (1-WL) test in distinguishing graphs in a graph isomorphism testing frame-work. To t...
[ "GNN", "k-WL", "expressiveness" ]
null
12,849
2210.09521
title_snapshot
[ -0.03415931388735771, -0.027371646836400032, 0.023670334368944168, 0.043252307921648026, 0.017162907868623734, 0.03081953153014183, 0.02536887675523758, 0.00833292119204998, -0.007118231616914272, -0.03565271943807602, 0.02023465558886528, -0.03051622398197651, -0.06366777420043945, 0.0316...
The Impact of Task Underspecification in Evaluating Deep Reinforcement Learning
https://openreview.net/forum?id=F_9w7Wl78IH
[ "Vindula Jayawardana", "Catherine H Tang", "Sirui Li", "Dajiang Suo", "Cathy Wu" ]
null
null
Evaluations of Deep Reinforcement Learning (DRL) methods are an integral part of scientific progress of the field. Beyond designing DRL methods for general intelligence, designing task-specific methods is becoming increasingly prominent for real-world applications. In these settings, the standard evaluation practice in...
[ "Reinforcement Learning", "Evaluation", "Scientific Progress", "Reliability", "Benchmarking" ]
Calls for a change in how performance evaluations are conducted in task-specific deep reinforcement learning and suggests using a family of MDPs instead of specific MDPs.
12,840
2210.08607
title_snapshot
[ -0.0390559621155262, -0.026883229613304138, -0.022949904203414917, 0.04459064453840256, 0.049721602350473404, 0.004408676642924547, 0.042741548269987106, 0.011954406276345253, -0.009193046018481255, -0.06170368939638138, -0.019122492522001266, 0.04619252681732178, -0.06903132796287537, -0....
Chaotic Dynamics are Intrinsic to Neural Network Training with SGD
https://openreview.net/forum?id=ffy-h0GKZbK
[ "Luis Herrmann", "Maximilian Granz", "Tim Landgraf" ]
null
null
With the advent of deep learning over the last decade, a considerable amount of effort has gone into better understanding and enhancing Stochastic Gradient Descent so as to improve the performance and stability of artificial neural network training. Active research fields in this area include exploiting second order in...
[ "Optimization", "Deep Learning", "Chaos", "Neural Networks", "Curvature", "Seconder Order Optimization", "SGD", "Hessian" ]
We find evidence that neural network training is intrinsically locally chaotic due to the negative eigenspectrum of the Hessian, and that network training exhibits globally edge-chaotic behaviour.
12,825
null
null
[ -0.048137158155441284, -0.03512315824627876, -0.01259678602218628, 0.02581249736249447, 0.032701071351766586, 0.032735660672187805, 0.028705308213829994, 0.007504552602767944, -0.04418278485536575, -0.058870065957307816, -0.01019006222486496, -0.02604076825082302, -0.03358578309416771, 0.0...
A PAC-Bayesian Generalization Bound for Equivariant Networks
https://openreview.net/forum?id=6dfYc2IUj4
[ "Arash Behboodi", "Gabriele Cesa", "Taco Cohen" ]
null
null
Equivariant networks capture the inductive bias about the symmetry of the learning task by building those symmetries into the model. In this paper, we study how equivariance relates to generalization error utilizing PAC Bayesian analysis for equivariant networks, where the transformation laws of feature spaces are dete...
[ "generalization error", "equivariant networks", "group representation", "PAC Bayesian" ]
We provide generalization bounds for equivariant networks in terms of group representations chosen for each layer.
12,818
2210.13150
title_snapshot
[ -0.0018660799833014607, 0.011134061962366104, 0.003535068593919277, 0.03441666439175606, 0.024707511067390442, 0.019330119714140892, 0.039438746869564056, -0.009090538136661053, -0.032036397606134415, -0.033922165632247925, 0.00003102223854511976, -0.012963182292878628, -0.07668942213058472,...
Autoregressive Perturbations for Data Poisoning
https://openreview.net/forum?id=1vusesyN7E
[ "Pedro Sandoval-Segura", "Vasu Singla", "Jonas Geiping", "Micah Goldblum", "Tom Goldstein", "David W. Jacobs" ]
null
null
The prevalence of data scraping from social media as a means to obtain datasets has led to growing concerns regarding unauthorized use of data. Data poisoning attacks have been proposed as a bulwark against scraping, as they make data ``unlearnable'' by adding small, imperceptible perturbations. Unfortunately, existing...
[ "autoregressive processes", "poisons", "data poisoning", "data protection", "imperceptible perturbations", "adversarial machine learning" ]
null
12,810
2206.03693
title_snapshot
[ 0.011210850439965725, -0.016564002260565758, -0.035078927874565125, 0.05018569901585579, 0.03906380757689476, 0.014972598291933537, 0.06365339457988739, -0.026693932712078094, -0.004992757458239794, -0.04475805535912514, -0.0038265965413302183, -0.00041421400965191424, -0.08613040298223495, ...
Near-Optimal No-Regret Learning Dynamics for General Convex Games
https://openreview.net/forum?id=SiSv_XDMksL
[ "Gabriele Farina", "Ioannis Anagnostides", "Haipeng Luo", "Chung-Wei Lee", "Christian Kroer", "Tuomas Sandholm" ]
null
null
A recent line of work has established uncoupled learning dynamics such that, when employed by all players in a game, each player's regret after $T$ repetitions grows polylogarithmically in $T$, an exponential improvement over the traditional guarantees within the no-regret framework. However, so far these results...
[ "No-regret learning", "optimism", "extensive-form games", "convex games" ]
We establish the first near-optimal learning dynamics for games with concave utilities and general convex strategy sets.
12,799
2206.08742
title_snapshot
[ -0.06442290544509888, -0.02496582455933094, -0.0016634793719276786, 0.029334932565689087, 0.043634314090013504, 0.033871471881866455, 0.001962076174095273, 0.047158386558294296, -0.03611258044838905, -0.04986611381173134, -0.014104830101132393, 0.0229193102568388, -0.05150775983929634, -0....
Learning the Structure of Large Networked Systems Obeying Conservation Laws
https://openreview.net/forum?id=WcxJooGBCc
[ "Anirudh Rayas", "Rajasekhar Anguluri", "Gautam Dasarathy" ]
null
null
Many networked systems such as electric networks, the brain, and social networks of opinion dynamics are known to obey conservation laws. Examples of this phenomenon include the Kirchoff laws in electric networks and opinion consensus in social networks. Conservation laws in networked systems are modeled as balance equ...
[ "Structure Learning", "Networked Systems", "Conservation Laws", "Gaussian Graphical models", "Sparsistency", "High dimensional regime" ]
null
12,795
2206.07083
title_snapshot
[ -0.028051123023033142, -0.02124211937189102, 0.026805533096194267, 0.03314518555998802, 0.01667875237762928, -0.0015931575326249003, -0.00039989768993109465, 0.0145179508253932, -0.015605198219418526, -0.033192917704582214, 0.021710503846406937, -0.025567805394530296, -0.06753137707710266, ...
Neural Payoff Machines: Predicting Fair and Stable Payoff Allocations Among Team Members
https://openreview.net/forum?id=CLMuNJSJfhv
[ "Daphne Cornelisse", "Thomas Rood", "Yoram Bachrach", "Mateusz Malinowski", "Tal Kachman" ]
null
null
In many multi-agent settings, participants can form teams to achieve collective outcomes that may far surpass their individual capabilities. Measuring the relative contributions of agents and allocating them shares of the reward that promote long-lasting cooperation are difficult tasks. Cooperative game theory offers s...
[ "Cooperative games theory", "Coalitional games", "Shapley values", "Bahnhof power index", "The Core" ]
Neural networks can approximately compute game theoretical solutions and speed up explainable AI methods
12,792
2208.08798
title_snapshot
[ -0.038759905844926834, -0.02128150314092636, -0.015075793489813805, 0.025791039690375328, 0.024597806856036186, 0.018545877188444138, -0.0032928306609392166, 0.014523898251354694, -0.019558561965823174, -0.051255203783512115, 0.0058138384483754635, 0.02045380137860775, -0.06443485617637634, ...
Implicit Neural Representations with Levels-of-Experts
https://openreview.net/forum?id=St5q10aqLTO
[ "Zekun Hao", "Arun Mallya", "Serge Belongie", "Ming-Yu Liu" ]
null
null
Coordinate-based networks, usually in the forms of MLPs, have been successfully applied to the task of predicting high-frequency but low-dimensional signals using coordinate inputs. To scale them to model large-scale signals, previous works resort to hybrid representations, combining a coordinate-based network with a g...
[ "Implicit neural representations", "neural fields", "coordinate-based networks", "hybrid representations", "positional encoding" ]
Coordinate-based MLP with periodic and multi-scale position-dependent weights arranged in multi-resolution grids.
12,785
null
null
[ -0.044764965772628784, -0.015080733224749565, 0.01628861203789711, 0.007799467071890831, 0.03035101667046547, 0.010821417905390263, -0.006066044792532921, 0.016818873584270477, -0.039626702666282654, -0.03304028883576393, 0.004950986709445715, -0.025046730414032936, -0.06046492978930473, 0...
LieGG: Studying Learned Lie Group Generators
https://openreview.net/forum?id=9sKZ60VtRmi
[ "Artem Moskalev", "Anna Sepliarskaia", "Ivan Sosnovik", "Arnold W.M. Smeulders" ]
null
null
Symmetries built into a neural network have appeared to be very beneficial for a wide range of tasks as it saves the data to learn them. We depart from the position that when symmetries are not built into a model a priori, it is advantageous for robust networks to learn symmetries directly from the data to fit a task f...
[ "invariance", "equivariance", "symmetry", "Lie groups", "interpretability" ]
null
12,774
2210.04345
title_snapshot
[ -0.03127972036600113, 0.0017945871222764254, 0.027461951598525047, 0.014880885370075703, 0.02351248450577259, 0.009522411972284317, 0.009987710043787956, -0.0024621118791401386, -0.030731171369552612, -0.030613508075475693, 0.013493513688445091, -0.02673986554145813, -0.058306608349084854, ...
Local Bayesian optimization via maximizing probability of descent
https://openreview.net/forum?id=YRDXX4IIA9
[ "Quan Nguyen", "Kaiwen Wu", "Jacob R. Gardner", "Roman Garnett" ]
null
null
Local optimization presents a promising approach to expensive, high-dimensional black-box optimization by sidestepping the need to globally explore the search space. For objective functions whose gradient cannot be evaluated directly, Bayesian optimization offers one solution -- we construct a probabilistic model of th...
[ "local optimization", "Bayesian optimization", "active learning" ]
We design a local Bayesian optimization policy that maximizes the probability of descending the objective function.
12,764
2210.11662
title_snapshot
[ -0.029075341299176216, 0.006297408603131771, 0.01565963216125965, 0.045204855501651764, 0.044868845492601395, 0.03858558461070061, 0.026886289939284325, -0.030744317919015884, -0.013847973197698593, -0.04411851242184639, -0.0032504848204553127, 0.011640077456831932, -0.05080828443169594, -...
A Closer Look at Learned Optimization: Stability, Robustness, and Inductive Biases
https://openreview.net/forum?id=cxZEBQFDoFK
[ "James Harrison", "Luke Metz", "Jascha Sohl-Dickstein" ]
null
null
Learned optimizers---neural networks that are trained to act as optimizers---have the potential to dramatically accelerate training of machine learning models. However, even when meta-trained across thousands of tasks at huge computational expense, blackbox learned optimizers often struggle with stability and generaliz...
[ "meta-learning", "learned optimization" ]
We investigate the stability properties of learned optimizers, and apply the insights gleaned to develop a learned optimization architecture that yields strong performance improvements over existing architectures.
12,754
2209.11208
title_snapshot
[ -0.034012965857982635, -0.01399767305701971, -0.0007469874108210206, 0.046006035059690475, 0.040232546627521515, 0.048027291893959045, 0.025453142821788788, 0.0038698625285178423, -0.04351901262998581, -0.038224153220653534, 0.0018778926460072398, 0.01145501434803009, -0.06751120835542679, ...
Empirical Gateaux Derivatives for Causal Inference
https://openreview.net/forum?id=8gUjpEsLCU
[ "Michael Jordan", "Yixin Wang", "Angela Zhou" ]
null
null
We study a constructive procedure that approximates Gateaux derivatives for statistical functionals by finite-differencing, with attention to causal inference functionals. We focus on the case where probability distributions are not known a priori but need also to be estimated from data, leading to empirical Gateaux de...
[ "causal inference", "double robustness", "bias-adjustment", "influence function", "semiparametric", "offline reinforcement learning" ]
null
12,748
null
null
[ -0.059879839420318604, -0.009459142573177814, -0.012615865096449852, 0.023763438686728477, 0.04353654384613037, 0.023688409477472305, 0.01891249045729637, -0.010811320506036282, -0.010288660414516926, -0.03622489422559738, 0.0070719411596655846, 0.029656285420060158, -0.0631747767329216, -...
Adaptive Interest for Emphatic Reinforcement Learning
https://openreview.net/forum?id=QTjJMy-UNO
[ "Martin Klissarov", "Rasool Fakoor", "Jonas Mueller", "Kavosh Asadi", "Taesup Kim", "Alex Smola" ]
null
null
Emphatic algorithms have shown great promise in stabilizing and improving reinforcement learning by selectively emphasizing the update rule. Although the emphasis fundamentally depends on an interest function which defines the intrinsic importance of each state, most approaches simply adopt a uniform interest over all ...
[ "emphatic temporal difference", "interest function", "meta gradients", "meta learning" ]
We propose a way to automatically learn the interest function of emphatic algorithms and verify our approach on a wide range of environments.
12,747
null
null
[ 0.002588263712823391, -0.03736003488302231, 0.012874254025518894, 0.03746066614985466, 0.0480986125767231, 0.04136212170124054, -0.006332641001790762, 0.026054449379444122, -0.025861091911792755, -0.029349463060498238, -0.005925614852458239, 0.018857017159461975, -0.06414775550365448, -0.0...
Human-Robotic Prosthesis as Collaborating Agents for Symmetrical Walking
https://openreview.net/forum?id=HBGvWy9Vxq
[ "Ruofan Wu", "Junmin Zhong", "Brent Abraham Wallace", "Xiang Gao", "He Huang", "Jennie Si" ]
null
null
This is the first attempt at considering human influence in the reinforcement learning control of a robotic lower limb prosthesis toward symmetrical walking in real world situations. We propose a collaborative multi-agent reinforcement learning (cMARL) solution framework for this highly complex and challenging human-pr...
[ "collaborative multi-agent reinforcement learning", "continuous control", "human-robotic prosthesis collaboration", "real world AI application" ]
A new automatic control design for a wearable robot by treating human and robotic prosthesis as collaborating agents toward symmetrical walking in real world situations
12,724
null
null
[ -0.008211899548768997, -0.025110291317105293, -0.025129523128271103, -0.007522452622652054, 0.0638495534658432, 0.01938464306294918, 0.021498795598745346, 0.03343707323074341, -0.035887524485588074, -0.05531308427453041, -0.013222722336649895, -0.017181621864438057, -0.05371470749378204, -...
Uni[MASK]: Unified Inference in Sequential Decision Problems
https://openreview.net/forum?id=GisHNaleWiA
[ "Micah Carroll", "Orr Paradise", "Jessy Lin", "Raluca Georgescu", "Mingfei Sun", "David Bignell", "Stephanie Milani", "Katja Hofmann", "Matthew Hausknecht", "Anca Dragan", "Sam Devlin" ]
null
null
Randomly masking and predicting word tokens has been a successful approach in pre-training language models for a variety of downstream tasks. In this work, we observe that the same idea also applies naturally to sequential decision making, where many well-studied tasks like behavior cloning, offline RL, inverse dynamic...
[ "Multi-task Learning", "Unsupervised Learning", "Reinforcement Learning", "Deep Learning" ]
We show how sequential decision making tasks can be thought of in terms of corresponding input maskings, enabling the training of a single model to perform all tasks at once.
12,723
2211.10869
title_judge
[ -0.007210673298686743, -0.009744733572006226, -0.02586609125137329, 0.047899648547172546, 0.026216845959424973, 0.02732723392546177, 0.053256336599588394, 0.04557417333126068, -0.0176902674138546, -0.0223634522408247, -0.02336926758289337, 0.025840196758508682, -0.0856580138206482, -0.0397...
Leveraging the Hints: Adaptive Bidding in Repeated First-Price Auctions
https://openreview.net/forum?id=hjqTeP05OMB
[ "Wei Zhang", "Yanjun Han", "Zhengyuan Zhou", "Aaron Flores", "Tsachy Weissman" ]
null
null
With the advent and increasing consolidation of e-commerce, digital advertising has very recently replaced traditional advertising as the main marketing force in the economy. In the past four years, a particularly important development in the digital advertising industry is the shift from second-price auctions to first...
[ "Online learning", "bandit", "first-price auction" ]
null
12,722
2211.06358
title_snapshot
[ -0.03140856698155403, -0.025382591411471367, -0.012612960301339626, 0.04327717050909996, 0.034314192831516266, 0.021529970690608025, 0.020394600927829742, 0.014163709245622158, -0.019678451120853424, -0.03476618230342865, -0.030697546899318695, 0.017780128866434097, -0.06632033735513687, -...
ReCo: Retrieve and Co-segment for Zero-shot Transfer
https://openreview.net/forum?id=8ViFz-5Mnnv
[ "Gyungin Shin", "Weidi Xie", "Samuel Albanie" ]
null
null
Semantic segmentation has a broad range of applications, but its real-world impact has been significantly limited by the prohibitive annotation costs necessary to enable deployment. Segmentation methods that forgo supervision can side-step these costs, but exhibit the inconvenient requirement to provide labelled exampl...
[ "semantic segmentation", "vision-language models", "image retrieval", "co-segmentation", "zero-shot transfer" ]
We propose a new framework for zero-shot transfer semantic segmentation, which retrieves a set of unlabelled images of a concept using a language-image pre-trained model and co-segments the category regions using modern image representations.
12,712
2206.07045
title_snapshot
[ -0.007895532995462418, -0.0400533564388752, -0.027738092467188835, 0.04586614668369293, 0.04848599433898926, 0.021012872457504272, 0.0041927192360162735, 0.0446215458214283, -0.005788651295006275, -0.029142815619707108, -0.05080102011561394, 0.0006264956318773329, -0.05500562861561775, 0.0...
Boosting the Performance of Generic Deep Neural Network Frameworks with Log-supermodular CRFs
https://openreview.net/forum?id=M_et7iOQC_s
[ "Hao Xiong", "Yangxiao Lu", "Nicholas Ruozzi" ]
null
null
Historically, conditional random fields (CRFs) were popular tools in a variety of application areas from computer vision to natural language processing, but due to their higher computational cost and weaker practical performance, they have, in many situations, fallen out of favor and been replaced by end-to-end deep ne...
[ "conditional random fields", "log-supermodular", "structured prediction" ]
A novel framework using log-supermodular conditional random fields (CRFs) to smooth/boost the performance of existing deep neural network models in a variety of domains.
12,697
null
null
[ -0.008115450851619244, -0.026960330083966255, 0.0027439012192189693, 0.05454258993268013, 0.048095691949129105, 0.028918545693159103, 0.0037102345377206802, -0.010398529469966888, -0.03552943840622902, -0.03457227349281311, -0.016383128240704536, 0.03896762803196907, -0.06651502102613449, ...
End-to-end Stochastic Optimization with Energy-based Model
https://openreview.net/forum?id=_sYOodxTMcF
[ "Lingkai Kong", "Jiaming Cui", "Yuchen Zhuang", "Rui Feng", "B. Aditya Prakash", "Chao Zhang" ]
null
null
Decision-focused learning (DFL) was recently proposed for stochastic optimization problems that involve unknown parameters. By integrating predictive modeling with an implicitly differentiable optimization layer, DFL has shown superior performance to the standard two-stage predict-then-optimize pipeline. However, most ...
[ "end-to-end stochastic optimization", "energy-based model", "decision-focused learning" ]
an new end-to-end stochastic optimization method with energy-based model
12,696
2211.13837
title_snapshot
[ -0.021477820351719856, -0.037198640406131744, 0.007085541728883982, 0.015982789918780327, 0.04108644649386406, 0.036092884838581085, 0.021837426349520683, 0.023963700979948044, -0.023356843739748, -0.03910963609814644, 0.005652347579598427, 0.01106136292219162, -0.06833826750516891, 0.0014...
EAGER: Asking and Answering Questions for Automatic Reward Shaping in Language-guided RL
https://openreview.net/forum?id=scfOjwTtZ8S
[ "Thomas Carta", "Pierre-Yves Oudeyer", "Olivier Sigaud", "sylvain lamprier" ]
null
null
Reinforcement learning (RL) in long horizon and sparse reward tasks is notoriously difficult and requires a lot of training steps. A standard solution to speed up the process is to leverage additional reward signals, shaping it to better guide the learning process. In the context of language-conditioned RL, the abstrac...
[ "language-conditioned RL", "automatic reward shaping", "intrinsic rewards", "exploration", "auxiliary objectives", "question generation", "question answering" ]
We propose an automated reward shaping method for guiding exploration in instruction following settings.
12,685
2206.09674
title_snapshot
[ -0.027903368696570396, -0.03426734730601311, -0.008220609277486801, 0.03769754618406296, 0.02866518869996071, 0.03949488326907158, 0.007980880327522755, -0.006161560770124197, -0.03175137937068939, -0.0027446295134723186, -0.06424553692340851, 0.07071469724178314, -0.05374607816338539, -0....
A Causal Analysis of Harm
https://openreview.net/forum?id=q9XPBhFgL6z
[ "Sander Beckers", "Hana Chockler", "Joseph Halpern" ]
null
null
As autonomous systems rapidly become ubiquitous, there is a growing need for a legal and regulatory framework to address when and how such a system harms someone. There have been several attempts within the philosophy literature to define harm, but none of them has proven capable of dealing with with the many examples ...
[ "harm", "causality", "utility" ]
We present a formal qualitative definition of harm that is based on contrastive causation and a default utility.
12,677
2210.05327
title_snapshot
[ 0.002746396232396364, -0.006173701956868172, -0.06036008894443512, 0.0007561861420981586, 0.04372132197022438, 0.005424274131655693, 0.0367656908929348, 0.007461600936949253, 0.0008950354531407356, -0.04254544526338577, -0.02683502808213234, 0.003693107981234789, -0.054963916540145874, -0....
On-Demand Sampling: Learning Optimally from Multiple Distributions
https://openreview.net/forum?id=FR289LMkmxZ
[ "Nika Haghtalab", "Michael Jordan", "Eric Zhao" ]
null
null
Societal and real-world considerations such as robustness, fairness, social welfare and multi-agent tradeoffs have given rise to multi-distribution learning paradigms, such as collaborative [Blum et al. 2017], group distributionally robust [Sagawa et al. 2019], and fair federated learning [Mohri et al. 2019]. In each o...
[ "Sample Complexity", "Distributionally Robust Optimization", "Collaborative Learning", "Learning Theory", "Minmax Equilibria", "Online Mirror Descent" ]
We give optimal sample complexity bounds for several multi-distribution learning problems using insights from finding min-max equilibria in stochastic zero-sum games.
12,671
2210.12529
title_snapshot
[ -0.014899042434990406, -0.009517201222479343, 0.023228304460644722, 0.04632062092423439, 0.04954640567302704, 0.029184041544795036, 0.012425960041582584, -0.005928210914134979, -0.0187598317861557, -0.037460941821336746, 0.005598118994385004, -0.011461819522082806, -0.08597486466169357, -0...
Logical Activation Functions: Logit-space equivalents of Probabilistic Boolean Operators
https://openreview.net/forum?id=m6HNNpQO8dc
[ "Scott C Lowe", "Robert Earle", "Jason d'Eon", "Thomas Trappenberg", "Sageev Oore" ]
null
null
The choice of activation functions and their motivation is a long-standing issue within the neural network community. Neuronal representations within artificial neural networks are commonly understood as logits, representing the log-odds score of presence of features within the stimulus. We derive logit-space operators...
[ "activation", "activation functions", "logit", "probabilistic", "Bayesian", "Boolean", "logic", "dendritic computation" ]
Propose new neural network activation functions based on logit-space equivalents of probabilistic Boolean logic operations (AND, OR, XNOR)
12,669
2110.11940
title_snapshot
[ -0.04233451187610626, 0.004380150232464075, -0.005789176560938358, 0.021717725321650505, 0.0567329116165638, 0.03348241373896599, 0.01770237274467945, -0.00001744908331602346, -0.037279751151800156, -0.034607332199811935, -0.02311854623258114, 0.01608937978744507, -0.07123913615942001, -0....
Dynamic pricing and assortment under a contextual MNL demand
https://openreview.net/forum?id=OptX3Db1P4
[ "Noemie Perivier", "Vineet Goyal" ]
null
null
We consider dynamic multi-product pricing and assortment problems under an unknown demand over T periods, where in each period, the seller decides on the price for each product or the assortment of products to offer to a customer who chooses according to an unknown Multinomial Logit Model (MNL). Such problems arise in ...
[ "dynamic pricing", "assortment optimization", "multinomial logit", "online learning", "contextual information" ]
null
12,668
2110.10018
title_snapshot
[ -0.01989973522722721, -0.015744028612971306, 0.003704680595546961, 0.05152870714664459, 0.03512635454535484, 0.0663401186466217, 0.001282799057662487, 0.024045690894126892, -0.03748999163508415, -0.021520137786865234, -0.040612198412418365, 0.013504236936569214, -0.05465507134795189, -0.03...
Off-Team Learning
https://openreview.net/forum?id=uOdTKkg2FtP
[ "Brandon Cui", "Hengyuan Hu", "Andrei Lupu", "Samuel Sokota", "Jakob Nicolaus Foerster" ]
null
null
Zero-shot coordination (ZSC) evaluates an algorithm by the performance of a team of agents that were trained independently under that algorithm. Off-belief learning (OBL) is a recent method that achieves state-of-the-art results in ZSC in the game Hanabi. However, the implementation of OBL relies on a belief model that...
[ "Multi-Agent Reinforcement Learning", "Reinforcement Learning", "Cooperative Multi-Agent Reinforcement Learning", "Deep Reinforcement Learning" ]
By training in an off-team manner, we can mitigate the training and testing time covariate shift of off-belief learning, resulting in near optimal zero-shot coordination and mitigate covariate shift in ad-hoc teamplay and proxy human-AI.
12,658
null
null
[ -0.03139756619930267, -0.021277721971273422, -0.03091624565422535, 0.011889203451573849, 0.01961962692439556, -0.0019940633792430162, 0.03290676325559616, 0.020319104194641113, -0.0221847053617239, -0.033632490783929825, -0.03362036123871803, 0.015361681580543518, -0.09204161167144775, -0....
A Deep Reinforcement Learning Framework for Column Generation
https://openreview.net/forum?id=zBlj0Cs6dw1
[ "Cheng Chi", "Amine Mohamed Aboussalah", "Elias Boutros Khalil", "Juyoung Wang", "Zoha Sherkat-Masoumi" ]
null
null
Column Generation (CG) is an iterative algorithm for solving linear programs (LPs) with an extremely large number of variables (columns). CG is the workhorse for tackling large-scale integer linear programs, which rely on CG to solve LP relaxations within a branch and bound algorithm. Two canonical applications are the...
[ "Reinforcement learning", "Column Generation", "Column selection", "Machine learning for optimization" ]
Reinforcement learning aided column selection in column generation
12,657
2206.02568
title_snapshot
[ -0.019736962392926216, -0.03381950780749321, 0.015860456973314285, 0.05194338411092758, 0.06125929206609726, 0.05379777401685715, 0.010417243465781212, 0.001985149923712015, -0.010334247723221779, -0.03138653188943863, -0.01175311952829361, -0.011037015356123447, -0.0889376848936081, 0.031...
Sublinear Algorithms for Hierarchical Clustering
https://openreview.net/forum?id=VPhhd5pv0Qs
[ "Arpit Agarwal", "Sanjeev Khanna", "Huan Li", "Prathamesh Patil" ]
null
null
Hierarchical clustering over graphs is a fundamental task in data mining and machine learning with applications in many domains including phylogenetics, social network analysis, and information retrieval. Specifically, we consider the recently popularized objective function for hierarchical clustering due to Dasgupta~\...
[ "hierarchical clustering", "clustering", "sublinear algorithms", "graph algorithms" ]
null
12,652
2206.07633
title_snapshot
[ -0.017072007060050964, -0.026101062074303627, 0.010563457384705544, 0.022488046437501907, 0.04190444201231003, 0.023014286532998085, 0.023589923977851868, 0.001259828801266849, -0.01035330630838871, -0.0502280592918396, -0.0005512293428182602, -0.03346294164657593, -0.07015645503997803, -0...
A Few Expert Queries Suffices for Sample-Efficient RL with Resets and Linear Value Approximation
https://openreview.net/forum?id=d19Dsqtw421
[ "Philip Amortila", "Nan Jiang", "Dhruv Madeka", "Dean Foster" ]
null
null
The current paper studies sample-efficient Reinforcement Learning (RL) in settings where only the optimal value function is assumed to be linearly-realizable. It has recently been understood that, even under this seemingly strong assumption and access to a generative model, worst-case sample complexities can be prohibi...
[ "Reinforcement learning", "imitation learning", "function approximation", "sample efficiency", "linear realizability" ]
While sample complexities in MDPs with linear optimal value functions can be exponentially large, we give a new method which shows that a surprisingly-little amount of expert advice permits sample efficiency.
12,643
2207.08342
title_snapshot
[ -0.01201722864061594, -0.025029733777046204, -0.004427745472639799, 0.0721983015537262, 0.04835075885057449, 0.015351744368672371, 0.02816866524517536, -0.018848199397325516, -0.026145095005631447, -0.028799688443541527, -0.033589694648981094, 0.02579290233552456, -0.07298032194375992, -0....
Certifying Some Distributional Fairness with Subpopulation Decomposition
https://openreview.net/forum?id=6mej19W1ppP
[ "Mintong Kang", "Linyi Li", "Maurice Weber", "Yang Liu", "Ce Zhang", "Bo Li" ]
null
null
Extensive efforts have been made to understand and improve the fairness of machine learning models based on observational metrics, especially in high-stakes domains such as medical insurance, education, and hiring decisions. However, there is a lack of certified fairness considering the end-to-end performance of an ML ...
[ "Certifying Fairness", "fairness constrained distribution", "distribution shifts" ]
We propose a general framework to certifying the distributional fairness of a trained model based on fairness constrained distribution.
12,641
2205.15494
title_snapshot
[ -0.01888005994260311, -0.021402712911367416, -0.014797190204262733, 0.04767761752009392, 0.04798804596066475, 0.028591645881533623, 0.0276383887976408, -0.02981412410736084, -0.01562526263296604, 0.00944491196423769, -0.0002504854928702116, 0.013421377167105675, -0.08673148602247238, -0.00...
Accelerating Certified Robustness Training via Knowledge Transfer
https://openreview.net/forum?id=QFMw21ZKaa_
[ "Pratik Vaishnavi", "Kevin Eykholt", "Amir Rahmati" ]
null
null
Training deep neural network classifiers that are certifiably robust against adversarial attacks is critical to ensuring the security and reliability of AI-controlled systems. Although numerous state-of-the-art certified training methods have been developed, they are computationally expensive and scale poorly with resp...
[ "Adversarial machine learning", "certified robustness", "randomized smoothing" ]
null
12,640
2210.14283
title_snapshot
[ 0.00024046051839832217, -0.04262404516339302, -0.016625501215457916, 0.06307757645845413, 0.05131419748067856, 0.00892565306276083, 0.024187322705984116, -0.02395113743841648, 0.007152728270739317, -0.018661368638277054, -0.015120510011911392, 0.015077760443091393, -0.04661773517727852, 0....
Fairness in Federated Learning via Core-Stability
https://openreview.net/forum?id=lKULHf7oFDo
[ "Bhaskar Ray Chaudhury", "Linyi Li", "Mintong Kang", "Bo Li", "Ruta Mehta" ]
null
null
Federated learning provides an effective paradigm to jointly optimize a model benefited from rich distributed data while protecting data privacy. Nonetheless, the heterogeneity nature of distributed data, especially in the non-IID setting, makes it challenging to define and ensure fairness among local agents. For insta...
[ "Fairness", "Federated Learning", "Core-Stability", "Social Choice" ]
We define the notion of core-stable fairness for federated learning with heterogeneous data, and propose CoreFed, an efficient FL protocol, to learn core-stable fair models.
12,631
2211.02091
title_snapshot
[ -0.044652339071035385, -0.04704464599490166, 0.00007838589954189956, 0.040870167315006256, 0.026652434840798378, 0.032508350908756256, -0.006937040947377682, -0.005240419413894415, -0.03853834792971611, -0.05292494222521782, -0.011481385678052902, -0.01171149779111147, -0.06440746039152145, ...
Learning NP-Hard Multi-Agent Assignment Planning using GNN: Inference on a Random Graph and Provable Auction-Fitted Q-learning
https://openreview.net/forum?id=VHzCiK727EL
[ "HYUNWOOK KANG", "Taehwan Kwon", "Jinkyoo Park", "James R. Morrison" ]
null
null
This paper explores the possibility of near-optimally solving multi-agent, multi-task NP-hard planning problems with time-dependent rewards using a learning-based algorithm. In particular, we consider a class of robot/machine scheduling problems called the multi-robot reward collection problem (MRRC). Such MRRC problem...
[ "Multi-agent assignment planning", "Reinforcement learning", "Graph neural network", "Mean-field inference" ]
Multi-agent sequential assignment learning with time-dependent rewards was addressed by combining Reinforcement learning theory, GNN theory and Auction theory.
12,603
1905.12204
title_snapshot
[ -0.044965293258428574, -0.024697689339518547, -0.027725299820303917, 0.05286884307861328, 0.03367258608341217, 0.04150327667593956, 0.004622808191925287, 0.01629442349076271, -0.01178312860429287, -0.030935870483517647, -0.015899688005447388, 0.017835214734077454, -0.0822247564792633, -0.0...
Memory safe computations with XLA compiler
https://openreview.net/forum?id=2S_GtHBtTUP
[ "Artem Artemev", "Yuze An", "Tilman Roeder", "Mark van der Wilk" ]
null
null
Software packages like TensorFlow and PyTorch are designed to support linear algebra operations, and their speed and usability determine their success. However, by prioritising speed, they often neglect memory requirements. As a consequence, the implementations of memory-intensive algorithms that are convenient in term...
[ "xla", "compiler", "gaussian processes", "sparse gaussian processes", "k-nearest neighbour" ]
The extension to the XLA compiler for automatic resolving memory overflows in machine learning programs. The impact of memory optimisations is demonstrated on sparse Gaussian processes.
12,595
2206.14148
title_snapshot
[ -0.029804574325680733, -0.013821333646774292, -0.02789449505507946, 0.002950521884486079, 0.026786042377352715, 0.019093051552772522, 0.00407799705862999, 0.007496561389416456, -0.01544989924877882, -0.028781628236174583, -0.011875568889081478, -0.012336025014519691, -0.06479467451572418, ...
A Communication-efficient Algorithm with Linear Convergence for Federated Minimax Learning
https://openreview.net/forum?id=TATzsweWfof
[ "Zhenyu Sun", "Ermin Wei" ]
null
null
In this paper, we study a large-scale multi-agent minimax optimization problem, which models many interesting applications in statistical learning and game theory, including Generative Adversarial Networks (GANs). The overall objective is a sum of agents' private local objective functions. We focus on the federated set...
[ "federated learning", "minimax optimization", "generalization bound" ]
We study federated minimax learning problem. We study its generalization performance through Rademacher complexity analysis and propose a novel communication-efficient algorithm that guarantees linear convergence to the optimal solution.
12,588
2206.01132
title_snapshot
[ -0.028445187956094742, -0.036833178251981735, 0.03173181787133217, 0.049200717359781265, 0.014086187817156315, 0.03997013717889786, 0.021550962701439857, -0.005000418517738581, 0.0018463097512722015, -0.060626234859228134, -0.004188284743577242, -0.014271209016442299, -0.07819405198097229, ...
On Efficient Online Imitation Learning via Classification
https://openreview.net/forum?id=h2imPVlCCyN
[ "Yichen Li", "Chicheng Zhang" ]
null
null
Imitation learning (IL) is a general learning paradigm for sequential decision-making problems. Interactive imitation learning, where learners can interactively query for expert annotations, has been shown to achieve provably superior sample efficiency guarantees compared with its offline counterpart or reinforcement l...
[ "Imitation Learning", "Online Learning", "Reinforcement Learning Theory" ]
We give new positive and negative computational and statistical results on the fundamental feasibility of regret minimization in online imitation learning with discrete action spaces, in the general nonrealizable case.
12,584
2209.12868
title_snapshot
[ -0.022526917979121208, -0.02065959759056568, 0.005559703800827265, 0.02976568602025509, 0.0386776439845562, 0.032340195029973984, 0.02338951826095581, 0.009979226626455784, -0.01665547862648964, -0.024677325040102005, -0.020718714222311974, 0.0019150691805407405, -0.07354985177516937, -0.0...
AMP: Automatically Finding Model Parallel Strategies with Heterogeneity Awareness
https://openreview.net/forum?id=VoLXWO1L-43
[ "Dacheng Li", "Hongyi Wang", "Eric Xing", "Hao Zhang" ]
null
null
Scaling up model sizes can lead to fundamentally new capabilities in many machine learning (ML) tasks. However, training big models requires strong distributed system expertise to carefully design model-parallel execution strategies that suit the model architectures and cluster setups. In this paper, we develop AMP, a ...
[ "Machine Learning Systems", "Model Parallelism", "Automation", "Heterogeneity" ]
We propose AMP, a framework that automatically derives optimal model parallelism strategies for heterogeneous environments.
12,583
2210.07297
title_snapshot
[ -0.021655147895216942, -0.038340404629707336, -0.009074371308088303, 0.022972002625465393, 0.030167127028107643, 0.034529510885477066, 0.01823853887617588, 0.00019683198479469866, -0.03357333689928055, -0.03516395017504692, 0.0217625480145216, -0.010847276076674461, -0.055062729865312576, ...
Nonstationary Dual Averaging and Online Fair Allocation
https://openreview.net/forum?id=8bk68fodvD5
[ "Luofeng Liao", "Yuan Gao", "Christian Kroer" ]
null
null
We consider the problem of fairly allocating sequentially arriving items to a set of individuals. For this problem, the recently-introduced PACE algorithm leverages the dual averaging algorithm to approximate competitive equilibria and thus generate online fair allocations. PACE is simple, distributed, and parameter-fr...
[ "online fair allocation", "Fisher markets", "fair division", "online convex optimization", "market equilibrium" ]
We provide the first online fair allocation guarantees for nonstationary input, and give new corresponding results for composite dual averaging.
12,580
2202.11614
title_snapshot
[ -0.03323959559202194, -0.028073564171791077, -0.010261554270982742, 0.016254020854830742, 0.014420824125409126, 0.027157491073012352, -0.013111929409205914, 0.024586306884884834, -0.04303060844540596, -0.044716816395521164, 0.008853072300553322, -0.04011616110801697, -0.09566525369882584, ...
New Definitions and Evaluations for Saliency Methods: Staying Intrinsic, Complete and Sound
https://openreview.net/forum?id=opw858PBJl6
[ "Arushi Gupta", "Nikunj Saunshi", "Dingli Yu", "Kaifeng Lyu", "Sanjeev Arora" ]
null
null
Saliency methods compute heat maps that highlight portions of an input that were most important for the label assigned to it by a deep net. Evaluations of saliency methods convert this heat map into a new masked input by retaining the $k$ highest-ranked pixels of the original input and replacing the rest with "uninform...
[ "saliency", "saliency methods", "saliency evaluation", "soundness", "sanity checks", "interpretability" ]
Inspired by the idea of soundness from logic systems, this paper provides a new dimension for intrinsic evaluations of saliency methods.
12,579
2211.02912
title_snapshot
[ -0.009297473356127739, -0.011962919495999813, 0.0004681968130171299, 0.042988020926713943, 0.016503987833857536, 0.026109211146831512, 0.009589805267751217, 0.010646020993590355, -0.03240584582090378, -0.0659187063574791, -0.0389728806912899, 0.02184661664068699, -0.04363415017724037, -0.0...
A Unified Framework for Deep Symbolic Regression
https://openreview.net/forum?id=2FNnBhwJsHK
[ "Mikel Landajuela", "Chak Lee", "Jiachen Yang", "Ruben Glatt", "Claudio P. Santiago", "Ignacio Aravena", "Terrell N. Mundhenk", "Garrett Mulcahy", "Brenden K. Petersen" ]
null
null
The last few years have witnessed a surge in methods for symbolic regression, from advances in traditional evolutionary approaches to novel deep learning-based systems. Individual works typically focus on advancing the state-of-the-art for one particular class of solution strategies, and there have been few attempts to...
[ "symbolic regression", "reinforcement learning", "combinatorial optimization" ]
We propose a strategy to integrate five disparate methods for symbolic regression into a unified framework, resulting in a new state-of-the-art on SRBench benchmarks.
12,575
null
null
[ -0.011429807171225548, -0.03493205085396767, -0.024676337838172913, 0.01427547913044691, 0.06930317729711533, 0.08132677525281906, 0.03644661605358124, -0.023544467985630035, -0.022889843210577965, -0.051869992166757584, -0.003062688745558262, 0.03755345195531845, -0.08277583122253418, 0.0...
Pitfalls of Epistemic Uncertainty Quantification through Loss Minimisation
https://openreview.net/forum?id=epjxT_ARZW5
[ "Viktor Bengs", "Eyke Hüllermeier", "Willem Waegeman" ]
null
null
Uncertainty quantification has received increasing attention in machine learning in the recent past. In particular, a distinction between aleatoric and epistemic uncertainty has been found useful in this regard. The latter refers to the learner's (lack of) knowledge and appears to be especially difficult to measure and...
[ "Uncertainty Quantification", "Empirical Loss Minimisation", "Proper Scoring Rules" ]
We show that recent approaches for epistemic uncertainty learning via minimizing a specific loss functions will in general not be faithful
12,571
2203.06102
title_snapshot
[ -0.006620101630687714, 0.0030653676949441433, -0.01314014196395874, 0.0328914150595665, 0.04061261564493179, 0.03884830325841904, 0.017000047490000725, -0.028573855757713318, -0.02298729307949543, -0.024028271436691284, -0.017766868695616722, 0.056267570704221725, -0.05834215134382248, -0....
Best of Both Worlds Model Selection
https://openreview.net/forum?id=9-vs8BucEoo
[ "Aldo Pacchiano", "Christoph Dann", "Claudio Gentile" ]
null
null
We study the problem of model selection in bandit scenarios in the presence of nested policy classes, with the goal of obtaining simultaneous adversarial and stochastic (``best of both worlds") high-probability regret guarantees. Our approach requires that each base learner comes with a candidate regret bound that may ...
[ "bandits", "linear bandits", "model selection", "policy classes", "best of both worlds", "reinforcement learning", "adversarial", "stochastic" ]
Model selection in bandit scenarios with best-of-both-worlds regret guarantees
12,565
2206.14912
title_snapshot
[ -0.054596979171037674, 0.007697688415646553, -0.01646520011126995, 0.05826088786125183, 0.029317662119865417, 0.01999678835272789, 0.026009468361735344, 0.0037582581862807274, -0.02227798104286194, -0.03210395202040672, -0.0021171430125832558, 0.015824956819415092, -0.060546740889549255, -...
Structuring Representations Using Group Invariants
https://openreview.net/forum?id=vWUmBjin_-o
[ "Mehran Shakerinava", "Arnab Kumar Mondal", "Siamak Ravanbakhsh" ]
null
null
A finite set of invariants can identify many interesting transformation groups. For example, distances, inner products and angles are preserved by Euclidean, Orthogonal and Conformal transformations, respectively. In an equivariant representation, the group invariants should remain constant on the embedding as we trans...
[ "Equivariance", "Invariance", "Geometry", "Group Theory", "Representation learning", "Self-supervised learning" ]
null
12,556
null
null
[ 0.0201505646109581, -0.0044988002628088, 0.010503832250833511, 0.03303835168480873, 0.013136924244463444, 0.024372834712266922, 0.02133287861943245, 0.005016162991523743, -0.023447424173355103, -0.021728528663516045, -0.039418358355760574, -0.017291439697146416, -0.07004760205745697, 0.016...
The Query Complexity of Cake Cutting
https://openreview.net/forum?id=u_7qyNFwkP8
[ "Simina Branzei", "Noam Nisan" ]
null
null
We consider the query complexity of cake cutting in the standard query model and give lower and upper bounds for computing approximately envy-free, perfect, and equitable allocations with the minimum number of cuts. The lower bounds are tight for computing contiguous envy-free allocations among $n=3$ players and for c...
[ "fair division", "cake cutting", "query complexity", "lower bounds", "upper bounds" ]
We consider the query complexity of cake cutting and show upper and lower bounds for finding approximately fair allocations (e.g. envy-free, perfect, equitable) in the standard query model for cake cutting.
12,549
1705.02946
title_snapshot
[ -0.04831628128886223, -0.011597554199397564, -0.01291156280785799, 0.027030987665057182, 0.04239383339881897, 0.01630357652902603, 0.0012728585861623287, -0.0009206493268720806, -0.03383027762174606, -0.026901613920927048, -0.028644802048802376, -0.0032335424330085516, -0.053070079535245895,...
Structural Pruning via Latency-Saliency Knapsack
https://openreview.net/forum?id=cUOR-_VsavA
[ "Maying Shen", "Hongxu Yin", "Pavlo Molchanov", "Lei Mao", "Jianna Liu", "Jose M. Alvarez" ]
null
null
Structural pruning can simplify network architecture and improve inference speed. We propose Hardware-Aware Latency Pruning (HALP) that formulates structural pruning as a global resource allocation optimization problem, aiming at maximizing the accuracy while constraining latency under a predefined budget on targeting ...
[ "model compression", "deep neural network pruning", "latency reduction" ]
null
12,544
2210.06659
title_snapshot
[ 0.0076498170383274555, -0.023114550858736038, -0.005170359741896391, 0.05223337188363075, 0.01854291372001171, 0.042964279651641846, 0.009362699463963509, 0.009662697091698647, -0.05997851863503456, -0.03673870861530304, -0.01089395210146904, -0.01715664565563202, -0.03368065133690834, -0....
Subgame Solving in Adversarial Team Games
https://openreview.net/forum?id=Roiw2Trm-qP
[ "Brian Hu Zhang", "Luca Carminati", "Federico Cacciamani", "Gabriele Farina", "Pierriccardo Olivieri", "Nicola Gatti", "Tuomas Sandholm" ]
null
null
In adversarial team games, a team of players sequentially faces a team of adversaries. These games are the simplest setting with multiple players where cooperation and competition coexist, and it is known that the information asymmetry among the team members makes equilibrium approximation computationally hard. Althoug...
[]
null
12,541
null
null
[ -0.0482604019343853, -0.02571035549044609, -0.0008999580750241876, 0.03491320461034775, 0.018283439800143242, 0.014657595194876194, 0.02154604159295559, -0.0021005754824727774, -0.021815100684762, -0.04664291813969612, -0.007136904168874025, -0.006368151865899563, -0.08520746231079102, -0....
Multi-Game Decision Transformers
https://openreview.net/forum?id=0gouO5saq6K
[ "Kuang-Huei Lee", "Ofir Nachum", "Sherry Yang", "Lisa Lee", "C. Daniel Freeman", "Sergio Guadarrama", "Ian Fischer", "Winnie Xu", "Eric Jang", "Henryk Michalewski", "Igor Mordatch" ]
null
null
A longstanding goal of the field of AI is a method for learning a highly capable, generalist agent from diverse experience. In the subfields of vision and language, this was largely achieved by scaling up transformer-based models and training them on large, diverse datasets. Motivated by this progress, we investigate w...
[ "Reinforcement Learning", "Generalist Agent", "Multi-Environment RL", "Upside-Down RL", "Decision Transformers" ]
We learn one Multi-Game Decision Transformer to achieve close to human-level performance on up to 41 Atari games.
12,529
2205.15241
title_snapshot
[ -0.03741881996393204, -0.027197908610105515, -0.003552928799763322, 0.02358146943151951, 0.03472406789660454, 0.02738376334309578, -0.0014132083160802722, 0.02044171467423439, -0.032578758895397186, -0.029435837641358376, -0.0252260472625494, 0.05072341114282608, -0.08615212142467499, -0.0...
Parameter-free Regret in High Probability with Heavy Tails
https://openreview.net/forum?id=fWHOcnHb1n
[ "Jiujia Zhang", "Ashok Cutkosky" ]
null
null
We present new algorithms for online convex optimization over unbounded domains that obtain parameter-free regret in high-probability given access only to potentially heavy-tailed subgradient estimates. Previous work in unbounded domains con- siders only in-expectation results for sub-exponential subgradients. Unlike i...
[ "Online learning", "Parameter-free", "Online Convex Optimization", "Heavy tails", "Regularization" ]
We produce parameter-free online learning algorithms whose regret bound holds in high probability even for heavy tailed subgradient estimates.
12,527
2210.14355
title_snapshot
[ -0.03394779562950134, 0.010644148103892803, 0.024701029062271118, 0.045265667140483856, 0.05012994259595871, 0.02657702937722206, 0.004788657184690237, 0.01971498318016529, 0.007628175429999828, -0.04201526567339897, -0.006741201039403677, -0.01275878120213747, -0.0585266537964344, -0.0233...
Learning to Compare Nodes in Branch and Bound with Graph Neural Networks
https://openreview.net/forum?id=0VhrZPJXcTU
[ "Abdel Ghani Labassi", "Didier Chételat", "Andrea Lodi" ]
null
null
Branch-and-bound approaches in integer programming require ordering portions of the space to explore next, a problem known as node comparison. We propose a new siamese graph neural network model to tackle this problem, where the nodes are represented as bipartite graphs with attributes. Similar to prior work, we train ...
[]
null
12,522
2210.16934
title_snapshot
[ -0.01774732582271099, -0.0240408256649971, -0.02921557053923607, 0.05970337614417076, 0.04638393223285675, 0.03493437543511391, 0.02553953230381012, -0.007133875507861376, -0.028712911531329155, -0.021901896223425865, 0.009742437861859798, -0.0048800865188241005, -0.08395551145076752, -0.0...
Communication Acceleration of Local Gradient Methods via an Accelerated Primal-Dual Algorithm with an Inexact Prox
https://openreview.net/forum?id=W72rB0wwLVu
[ "Abdurakhmon Sadiev", "Dmitry Kovalev", "Peter Richtárik" ]
null
null
Inspired by a recent breakthrough of Mishchenko et al. [2022], who for the first time showed that local gradient steps can lead to provable communication acceleration, we propose an alternative algorithm which obtains the same communication acceleration as their method (ProxSkip). Our approach is very different, howeve...
[ "ProxSkip", "Communication Acceleration", "Federated Learning", "Local Gradient Descent", "Federated Averaging", "Primal-Dual Methods" ]
null
12,515
2207.03957
title_judge
[ -0.040009383112192154, -0.05835650861263275, 0.006608873605728149, 0.048095282167196274, 0.025482511147856712, 0.04133303463459015, 0.027222488075494766, -0.00519292987883091, 0.0015268151182681322, -0.062334682792425156, 0.005596595350652933, 0.0021438398398458958, -0.0581839419901371, -0...
On the detrimental effect of invariances in the likelihood for variational inference
https://openreview.net/forum?id=ft4xGJ8tIZH
[ "Richard Kurle", "Ralf Herbrich", "Tim Januschowski", "Bernie Wang", "Jan Gasthaus" ]
null
null
Variational Bayesian posterior inference often requires simplifying approximations such as mean-field parametrisation to ensure tractability. However, prior work has associated the variational mean-field approximation for Bayesian neural networks with underfitting in the case of small datasets or large model sizes. In ...
[ "Bayesian Neural Networks", "Variational Bayes", "Variational Inference", "Invariance", "Symmetry", "posterior collapse" ]
Invariances in the likelihood of over-parametrised models such as neural networks complicate the structure of the posterior by introducing additional modes that can not be approximated by mean-field distributions.
12,506
2209.07157
title_snapshot
[ -0.013617003336548805, 0.011743358336389065, -0.0024045747704803944, 0.01708797551691532, 0.026016665622591972, 0.03977969288825989, 0.04713957756757736, -0.005650754552334547, -0.034042857587337494, -0.03642725571990013, -0.007729847449809313, 0.020359033718705177, -0.06331898272037506, -...
BayesPCN: A Continually Learnable Predictive Coding Associative Memory
https://openreview.net/forum?id=9cPDqh9fQMy
[ "Jinsoo Yoo", "Frank Wood" ]
null
null
Associative memory plays an important role in human intelligence and its mechanisms have been linked to attention in machine learning. While the machine learning community's interest in associative memories has recently been rekindled, most work has focused on memory recall ($read$) over memory learning ($write$). In t...
[ "machine learning", "associative memory", "predictive coding", "continual learning", "bayesian inference" ]
We continually learn a hierarchical assocative memory that can recall i.i.d. high-dimensional data observed hundreds of timesteps ago using predictive coding and locally conjugate Bayesian updates.
12,488
2205.09930
title_snapshot
[ -0.024751493707299232, 0.01446594949811697, -0.00006095505887060426, 0.027349574491381645, 0.024527210742235184, 0.02882847748696804, 0.0011679610470309854, 0.013110456988215446, -0.0508740097284317, -0.02866096794605255, -0.006874439772218466, -0.0005787993432022631, -0.05294713377952576, ...
Robustness to Unbounded Smoothness of Generalized SignSGD
https://openreview.net/forum?id=8oj_2Ypp0j
[ "Michael Crawshaw", "Mingrui Liu", "Francesco Orabona", "Wei Zhang", "Zhenxun Zhuang" ]
null
null
Traditional analyses in non-convex optimization typically rely on the smoothness assumption, namely requiring the gradients to be Lipschitz. However, recent evidence shows that this smoothness condition does not capture the properties of some deep learning objective functions, including the ones involving Recurrent Neu...
[ "stochastic", "optimization", "noncovex", "adam", "transformer", "clipping", "signsgd", "momentum", "unbounded smoothness" ]
We analyze a generalized signsgd with a relaxed smoothness assumption that is verified in practice in Transformer training.
12,485
2208.11195
title_snapshot
[ -0.04313221201300621, -0.02725144661962986, 0.031618185341358185, 0.03247564285993576, 0.038938168436288834, 0.04304179176688194, 0.03521633520722389, 0.029332423582673073, -0.007607114966958761, -0.049160826951265335, -0.007510023191571236, 0.0037205561529845, -0.0685850977897644, -0.0169...
Generalization for multiclass classification with overparameterized linear models
https://openreview.net/forum?id=ikWvMRVQBWW
[ "Vignesh Subramanian", "Rahul Arya", "Anant Sahai" ]
null
null
Via an overparameterized linear model with Gaussian features, we provide conditions for good generalization for multiclass classification of minimum-norm interpolating solutions in an asymptotic setting where both the number of underlying features and the number of classes scale with the number of training points. The ...
[ "overparameterized", "multiclass", "classification", "theory", "generalization", "interpolation", "bi-level", "Gaussian model" ]
Asymptotic region for multiclass classification in overparameterized Gaussian feature model learned by min-norm interpolation where the number of classes scales with the number of training points.
12,480
2206.01399
title_snapshot
[ -0.0035552168264985085, -0.024846116080880165, 0.0070648835971951485, 0.02746673859655857, 0.031325481832027435, 0.0515611469745636, 0.022835416719317436, -0.01605863682925701, -0.02716396376490593, -0.03703193739056587, -0.028092680498957634, 0.019573593512177467, -0.09171221405267715, 0....
Finite-Sample Maximum Likelihood Estimation of Location
https://openreview.net/forum?id=1l5hEEK_j13
[ "Shivam Gupta", "Jasper C.H. Lee", "Eric Price", "Paul Valiant" ]
null
null
We consider 1-dimensional location estimation, where we estimate a parameter $\lambda$ from $n$ samples $\lambda + \eta_i$, with each $\eta_i$ drawn i.i.d. from a known distribution $f$. For fixed $f$ the maximum-likelihood estimate (MLE) is well-known to be optimal in the limit as $n \to \infty$: it is asymptotically ...
[]
null
12,479
2206.02348
title_snapshot
[ -0.0238932017236948, -0.0074170976877212524, 0.026350215077400208, 0.0009659416973590851, 0.061066195368766785, 0.018900593742728233, 0.021504772827029228, 0.03387099876999855, -0.04141636937856674, -0.0520089715719223, 0.03149060159921646, -0.03500988706946373, -0.05061940848827362, -0.01...
Graphein - a Python Library for Geometric Deep Learning and Network Analysis on Biomolecular Structures and Interaction Networks
https://openreview.net/forum?id=9xRZlV6GfOX
[ "Arian Rokkum Jamasb", "Ramon Viñas Torné", "Eric J Ma", "Yuanqi Du", "Charles Harris", "Kexin Huang", "Dominic Hall", "Pietro Lio", "Tom Leon Blundell" ]
null
null
Geometric deep learning has broad applications in biology, a domain where relational structure in data is often intrinsic to modelling the underlying phenomena. Currently, efforts in both geometric deep learning and, more broadly, deep learning applied to biomolecular tasks have been hampered by a scarcity of appropri...
[ "Protein", "Drug Discovery", "Geometric Deep Learning", "Biomolecules" ]
Python library for working with geometric representations of biomolecular structures
12,458
null
null
[ -0.012229016050696373, -0.03096316009759903, -0.03783637285232544, 0.027057301253080368, 0.011396369896829128, -0.0013628334272652864, 0.00966076273471117, 0.018825458362698555, 0.004108004737645388, -0.05739981308579445, 0.023050865158438683, -0.02210504375398159, -0.08909203857183456, 0....
Root Cause Analysis of Failures in Microservices through Causal Discovery
https://openreview.net/forum?id=weoLjoYFvXY
[ "Muhammad Azam Ikram", "Sarthak Chakraborty", "Subrata Mitra", "Shiv Saini", "Saurabh Bagchi", "Murat Kocaoglu" ]
null
null
Most cloud applications use a large number of smaller sub-components (called microservices) that interact with each other in the form of a complex graph to provide the overall functionality to the user. While the modularity of the microservice architecture is beneficial for rapid software development, maintaining and d...
[ "root cause analysis", "causal discovery", "causal graphs" ]
We create a solution for root cause diagnosis in microservice-based cloud computing applications by a scalable intervention-based causal discovery algorithm.
12,457
null
null
[ -0.014834705740213394, -0.02964947372674942, -0.015068708918988705, 0.05706458166241646, 0.0511774905025959, 0.0392075814306736, 0.040498510003089905, 0.015316893346607685, -0.027363207191228867, -0.03282683342695236, -0.0049820528365671635, -0.014466006308794022, -0.04501355066895485, 0.0...
Robust Model Selection and Nearly-Proper Learning for GMMs
https://openreview.net/forum?id=JCbLxJ1E6SO
[ "Allen Liu", "Jerry Li", "Ankur Moitra" ]
null
null
In learning theory, a standard assumption is that the data is generated from a finite mixture model. But what happens when the number of components is not known in advance? The problem of estimating the number of components, also called model selection, is important in its own right but there are essentially no known e...
[ "Mixtures of Gaussians", "model selection", "proper learning", "density estimation" ]
We give efficient algorithms for robust model selection and nearly-proper learning of Gaussian mixture models.
12,456
2106.02774
title_snapshot
[ -0.03270125389099121, -0.011693855747580528, 0.016857344657182693, 0.022437529638409615, 0.03404781594872475, 0.04666044935584068, 0.029120411723852158, 0.031535759568214417, -0.05707237496972084, -0.033099036663770676, -0.003318520961329341, 0.0304669588804245, -0.04692690819501877, -0.00...
Explain My Surprise: Learning Efficient Long-Term Memory by predicting uncertain outcomes
https://openreview.net/forum?id=5wdvW_hI7bP
[ "Artyom Sorokin", "Nazar Buzun", "Leonid Pugachev", "Mikhail Burtsev" ]
null
null
In many sequential tasks, a model needs to remember relevant events from the distant past to make correct predictions. Unfortunately, a straightforward application of gradient based training requires intermediate computations to be stored for every element of a sequence. This requires to store prohibitively large inter...
[ "Memory", "RNN", "Information Theory", "Reinforcement Learning", "POMDP" ]
null
12,446
2207.13649
title_snapshot
[ -0.028101306408643723, -0.03245813027024269, -0.015629583969712257, 0.034978944808244705, 0.045551761984825134, 0.027091531082987785, 0.019560987129807472, 0.018780339509248734, -0.043208345770835876, -0.02325408346951008, -0.0049483138136565685, 0.009191476739943027, -0.060010287910699844, ...
Structural Knowledge Distillation for Object Detection
https://openreview.net/forum?id=O3My0RK9s_R
[ "Philip De Rijk", "Lukas Schneider", "Marius Cordts", "Dariu Gavrila" ]
null
null
Knowledge Distillation (KD) is a well-known training paradigm in deep neural networks where knowledge acquired by a large teacher model is transferred to a small student. KD has proven to be an effective technique to significantly improve the student's performance for various tasks including object detection. As such,...
[ "computer vision", "CNNs", "knowledge distillation", "object detection", "deep learning" ]
State-of-the-art knowledge distillation performance for object detectors using structural similarity
12,426
2211.13133
title_snapshot
[ 0.017067590728402138, -0.00464244931936264, 0.008684275671839714, 0.05873057246208191, 0.04337834566831589, 0.005349259823560715, 0.03509678319096565, -0.03034757636487484, -0.022735238075256348, -0.011589068919420242, -0.03984406590461731, -0.013213064521551132, -0.05023609846830368, 0.00...
Exploring the Latent Space of Autoencoders with Interventional Assays
https://openreview.net/forum?id=hdZeYGNCTtN
[ "Felix Leeb", "Stefan Bauer", "Michel Besserve", "Bernhard Schölkopf" ]
null
null
Autoencoders exhibit impressive abilities to embed the data manifold into a low-dimensional latent space, making them a staple of representation learning methods. However, without explicit supervision, which is often unavailable, the representation is usually uninterpretable, making analysis and principled progress cha...
[ "Autoencoders", "Interventions", "Structured Representation", "Manifold Learning" ]
We develop tools to better understand the learned structure of autoencoders based on self-consistency to improve performance on downstream tasks.
12,410
2106.16091
title_snapshot
[ 0.008713951334357262, -0.04630178585648537, -0.038431018590927124, 0.045534875243902206, 0.028106000274419785, 0.04252203181385994, 0.04063621163368225, -0.011678487993776798, -0.013487430289387703, -0.03767988085746765, 0.00414531072601676, -0.008467365987598896, -0.06523041427135468, 0.0...
Performative Power
https://openreview.net/forum?id=NqDXfe2oC_1
[ "Moritz Hardt", "Meena Jagadeesan", "Celestine Mendler-Dünner" ]
null
null
We introduce the notion of performative power, which measures the ability of a firm operating an algorithmic system, such as a digital content recommendation platform, to cause change in a population of participants. We relate performative power to the economic study of competition in digital economies. Traditional eco...
[ "markets", "competition", "power", "performativity", "prediction" ]
We introduce the notion of performative power that quantifies a firm’s ability to steer a population of participants in a digital economy
12,409
2203.17232
title_snapshot
[ -0.017127562314271927, -0.027864398434758186, -0.017896926030516624, 0.03330947831273079, 0.0651383250951767, -0.0006974752759560943, 0.013902740553021431, 0.03571872040629387, -0.010676992125809193, -0.004241368733346462, -0.011862361803650856, -0.03208814933896065, -0.041960459202528, -0...
Debiased Machine Learning without Sample-Splitting for Stable Estimators
https://openreview.net/forum?id=anqloMQdWtP
[ "Qizhao Chen", "Vasilis Syrgkanis", "Morgane Austern" ]
null
null
Estimation and inference on causal parameters is typically reduced to a generalized method of moments problem, which involves auxiliary functions that correspond to solutions to a regression or classification problem. Recent line of work on debiased machine learning shows how one can use generic machine learning estima...
[ "Double Machine Learning", "Stability", "Causal Inference", "Treatment Effects", "Debiased Machine Learning" ]
We prove asymptotic normality for a target parameter of interest, of debiased machine learning semi-parametric estimators without sample splitting, when the machine learning estimators used for the nuisance functions are leave-one-out stable.
12,400
2206.01825
title_snapshot
[ -0.012181554920971394, -0.03349899500608444, -0.028358856216073036, 0.01987580582499504, 0.03146982938051224, 0.0391140952706337, 0.027786068618297577, -0.03792085126042366, -0.024196861311793327, -0.05351835861802101, -0.007515807170420885, 0.025277195498347282, -0.10049112141132355, -0.0...
HyperTree Proof Search for Neural Theorem Proving
https://openreview.net/forum?id=J4pX8Q8cxHH
[ "Guillaume Lample", "Timothee Lacroix", "Marie-anne Lachaux", "Aurelien Rodriguez", "Amaury Hayat", "Thibaut Lavril", "Gabriel Ebner", "Xavier Martinet" ]
null
null
We propose an online training procedure for a transformer-based automated theorem prover. Our approach leverages a new search algorithm, HyperTree Proof Search (HTPS), that learns from previous proof searches through online training, allowing it to generalize to domains far from the training distribution. We report det...
[ "theorem proving", "automated theorem proving", "MCTS", "reasoning", "AI for math" ]
An AlphaZero-inspired method for automated theorem proving.
12,399
2205.11491
title_snapshot
[ -0.02024165913462639, -0.0017262627370655537, 0.016894251108169556, 0.06017705798149109, 0.02088235504925251, 0.01710125245153904, 0.032247573137283325, -0.03914029896259308, -0.01483934000134468, -0.013879183679819107, -0.0048085604794323444, 0.024965308606624603, -0.03419173136353493, 0....
Spherical Channels for Modeling Atomic Interactions
https://openreview.net/forum?id=5Z3GURcqwT
[ "C. Lawrence Zitnick", "Abhishek Das", "Adeesh Kolluru", "Janice Lan", "Muhammed Shuaibi", "Anuroop Sriram", "Zachary Ward Ulissi", "Brandon M Wood" ]
null
null
Modeling the energy and forces of atomic systems is a fundamental problem in computational chemistry with the potential to help address many of the world’s most pressing problems, including those related to energy scarcity and climate change. These calculations are traditionally performed using Density Functional Theor...
[ "graph neural networks", "equivariance", "invariance", "materials science", "chemistry" ]
A graph neural network using spherical function channels that relaxes the rotational equivariance constraint for modeling atomic interactions.
12,396
2206.14331
title_snapshot
[ -0.027626793831586838, 0.01469469629228115, 0.0006995780277065933, 0.03611520305275917, 0.02855445258319378, -0.012453797273337841, -0.0012740247184410691, 0.0015064295148476958, -0.0004857074236497283, -0.05300333723425865, 0.03700779005885124, -0.0012570797698572278, -0.07283351570367813, ...
Chroma-VAE: Mitigating Shortcut Learning with Generative Classifiers
https://openreview.net/forum?id=WWVcsfI0jGH
[ "Wanqian Yang", "Polina Kirichenko", "Micah Goldblum", "Andrew Gordon Wilson" ]
null
null
Deep neural networks are susceptible to shortcut learning, using simple features to achieve low training loss without discovering essential semantic structure. Contrary to prior belief, we show that generative models alone are not sufficient to prevent shortcut learning, despite an incentive to recover a more comprehen...
[ "deep generative model", "variational autoencoder", "generative classifier", "shortcut learning", "spurious correlation", "simplicity bias" ]
We use generative classifiers to tackle shortcut learning.
12,386
2211.15231
title_snapshot
[ 0.02122930809855461, -0.0021529938094317913, -0.03320692852139473, 0.048460498452186584, 0.012785731814801693, 0.03735553100705147, 0.04094770923256874, -0.0038999987300485373, -0.03812570124864578, -0.03663410618901253, -0.04398776590824127, 0.0364689975976944, -0.05644071102142334, 0.005...
VisCo Grids: Surface Reconstruction with Viscosity and Coarea Grids
https://openreview.net/forum?id=tX_dIvk4j-s
[ "Albert Pumarola", "Artsiom Sanakoyeu", "Lior Yariv", "Ali Thabet", "Yaron Lipman" ]
null
null
Surface reconstruction has been seeing a lot of progress lately by utilizing Implicit Neural Representations (INRs). Despite their success, INRs often introduce hard to control inductive bias (i.e., the solution surface can exhibit unexplainable behaviours), have costly inference, and are slow to train. The goal of th...
[ "surface reconstruction", "implicit neural representations", "signed distance functions" ]
We introduce VisCo, a grid-based surface reconstruction method incorporating viscosity and coarea priors.
12,378
2303.14569
title_snapshot
[ -0.043025821447372437, -0.010496417060494423, 0.03120988793671131, 0.047748226672410965, 0.024109071120619774, 0.04023173823952675, -0.0027012480422854424, 0.021435456350445747, -0.028531858697533607, -0.07115430384874344, 0.01616029255092144, -0.026064306497573853, -0.037576958537101746, ...
Learning Probabilistic Models from Generator Latent Spaces with Hat EBM
https://openreview.net/forum?id=AluQNIIb_Zy
[ "Mitch Hill", "Erik Nijkamp", "Jonathan Craig Mitchell", "Bo Pang", "Song-Chun Zhu" ]
null
null
This work proposes a method for using any generator network as the foundation of an Energy-Based Model (EBM). Our formulation posits that observed images are the sum of unobserved latent variables passed through the generator network and a residual random variable that spans the gap between the generator output and the...
[ "EBM", "latent sampling", "generator network", "MCMC", "Langevin" ]
We propose the Hat EBM to learn probabilistic models that can incorporate any kind of generator network.
12,375
2210.16486
title_snapshot
[ 0.027814384549856186, -0.02647397294640541, -0.0025713369250297546, 0.039068881422281265, 0.03256211429834366, 0.007675749249756336, 0.0020313675049692392, -0.005823869723826647, -0.03728242591023445, -0.0417623296380043, 0.008179919794201851, -0.0026435828767716885, -0.06810417771339417, ...
On the Effectiveness of Fine-tuning Versus Meta-reinforcement Learning
https://openreview.net/forum?id=mux7gn3g_3
[ "Zhao Mandi", "Pieter Abbeel", "Stephen James" ]
null
null
Intelligent agents should have the ability to leverage knowledge from previously learned tasks in order to learn new ones quickly and efficiently. Meta-learning approaches have emerged as a popular solution to achieve this. However, meta-reinforcement learning (meta-RL) algorithms have thus far been restricted to simpl...
[ "Reinforcement Learning", "Fine-tuning", "Multi-task Learning", "Meta Reinforcement Learning", "Meta-RL" ]
Multi-task pretraining followed by fine-tuning on novel tasks performs equally as well, or better, than common meta-RL baselines.
12,373
2206.03271
title_snapshot
[ -0.014432278461754322, -0.02370877005159855, 0.008394289761781693, 0.034056492149829865, 0.05036862567067146, 0.008919328451156616, 0.026236847043037415, 0.009779475629329681, -0.03705475106835365, -0.03228059411048889, -0.03064415417611599, 0.06589320302009583, -0.06759633123874664, -0.02...
SatMAE: Pre-training Transformers for Temporal and Multi-Spectral Satellite Imagery
https://openreview.net/forum?id=WBhqzpF6KYH
[ "Yezhen Cong", "Samar Khanna", "Chenlin Meng", "Patrick Liu", "Erik Rozi", "Yutong He", "Marshall Burke", "David B. Lobell", "Stefano Ermon" ]
null
null
Unsupervised pre-training methods for large vision models have shown to enhance performance on downstream supervised tasks. Developing similar techniques for satellite imagery presents significant opportunities as unlabelled data is plentiful and the inherent temporal and multi-spectral structure provides avenues to fu...
[ "Self-supervised learning", "Masked Autoencoder", "Satellite imagery" ]
In this paper, we present SatMAE, a new SOTA pre-training framework for temporal or multi-spectral satellite imagery based on Masked Autoencoder (MAE).
12,368
2207.08051
title_snapshot
[ 0.03409763053059578, -0.048040326684713364, -0.019084110856056213, 0.02882324531674385, 0.03672702610492706, 0.03939250111579895, 0.01899862475693226, 0.020207924768328667, -0.040841542184352875, -0.0254451185464859, -0.06398845463991165, 0.02188648097217083, -0.06274358928203583, 0.001205...
Maximizing and Satisficing in Multi-armed Bandits with Graph Information
https://openreview.net/forum?id=KQYodS0W0j
[ "Parth Kashyap Thaker", "Mohit Malu", "Nikhil Rao", "Gautam Dasarathy" ]
null
null
Pure exploration in multi-armed bandits has emerged as an important framework for modeling decision making and search under uncertainty. In modern applications however, one is often faced with a tremendously large number of options and even obtaining one observation per option may be too costly rendering traditional pu...
[ "Multi-armed Bandits", "Pure Exploration", "Graph Smoothness", "Best Arm Identification", "Sample Complexity" ]
null
12,365
2108.01152
title_snapshot
[ -0.03167998418211937, -0.011255494318902493, 0.005039633251726627, 0.0347108356654644, 0.04302610829472542, 0.03384692594408989, 0.033625781536102295, 0.015292570926249027, -0.013202953152358532, -0.055609025061130524, -0.011396297253668308, 0.009223142638802528, -0.05873306840658188, -0.0...
$k$-Sliced Mutual Information: A Quantitative Study of Scalability with Dimension
https://openreview.net/forum?id=L-ceBdl2DPb
[ "Ziv Goldfeld", "Kristjan Greenewald", "Theshani Nuradha", "Galen Reeves" ]
null
null
Sliced mutual information (SMI) is defined as an average of mutual information (MI) terms between one-dimensional random projections of the random variables. It serves as a surrogate measure of dependence to classic MI that preserves many of its properties but is more scalable to high dimensions. However, a quantitativ...
[]
null
12,358
2206.08526
title_snapshot
[ -0.04427558556199074, -0.03298084810376167, 0.02766227349638939, 0.025590434670448303, 0.03719932213425636, 0.016437608748674393, 0.0369621142745018, -0.012243079021573067, -0.02536536194384098, -0.030390089377760887, 0.0011572570074349642, -0.004367327317595482, -0.05310855433344841, 0.00...
Distinguishing discrete and continuous behavioral variability using warped autoregressive HMMs
https://openreview.net/forum?id=6Kj1wCgiUp_
[ "Julia C Costacurta", "Lea Duncker", "Blue Sheffer", "Winthrop Gillis", "Caleb Weinreb", "Jeffrey Evan Markowitz", "Sandeep R. Datta", "Alex H Williams", "Scott Linderman" ]
null
null
A core goal in systems neuroscience and neuroethology is to understand how neural circuits generate naturalistic behavior. One foundational idea is that complex naturalistic behavior may be composed of sequences of stereotyped behavioral syllables, which combine to generate rich sequences of actions. To investigate thi...
[ "naturalistic behavior", "clustering", "markov models", "time series" ]
Warped autoregressive HMMs disentangle movement type and speed in behavioral recordings.
12,353
null
null
[ -0.0022450697142630816, 0.020762039348483086, -0.055154502391815186, 0.028721842914819717, 0.03979755938053131, 0.03477044403553009, 0.08838415145874023, 0.021097127348184586, -0.05456029251217842, -0.046675749123096466, -0.0062224105931818485, 0.003845152212306857, -0.05134119838476181, -...
Analyzing Data-Centric Properties for Graph Contrastive Learning
https://openreview.net/forum?id=7-LTDcvNc_
[ "Puja Trivedi", "Ekdeep Singh Lubana", "Mark Heimann", "Danai Koutra", "Jayaraman J. Thiagarajan" ]
null
null
Recent analyses of self-supervised learning (SSL) find the following data-centric properties to be critical for learning good representations: invariance to task-irrelevant semantics, separability of classes in some latent space, and recoverability of labels from augmented samples. However, given their discrete, non-Eu...
[ "unsupervised representation learning", "generalization", "graph neural networks", "augmentation", "invariance", "contrastive learning", "self supervised learning" ]
We study the behavior, theoretically and empirically, of graph contrastive learning with respect to data-dependent properties (invariance, recoverability and separability).
12,349
2208.02810
title_snapshot
[ -0.0027494204696267843, -0.032140083611011505, -0.006047843489795923, 0.05146244540810585, 0.025061780586838722, 0.003098846413195133, 0.05175693333148956, 0.012584848329424858, -0.009061651304364204, -0.03645385801792145, -0.015823230147361755, -0.004658416844904423, -0.07270068675279617, ...
Combining Explicit and Implicit Regularization for Efficient Learning in Deep Networks
https://openreview.net/forum?id=sADLRl2STMe
[ "Dan Zhao" ]
null
null
Works on implicit regularization have studied gradient trajectories during the optimization process to explain why deep networks favor certain kinds of solutions over others. In deep linear networks, it has been shown that gradient descent implicitly regularizes toward low-rank solutions on matrix completion/factorizat...
[ "optimization", "regularization", "deep learning", "matrix factorization", "neural networks", "efficient learning", "implicit regularization", "explicit regularization", "Matrix Completion", "Optimization for Deep Networks" ]
Explicit regluarization and optimizer's own inductive biases can combine to form beneficial effects for efficient learning
12,345
2306.00342
title_snapshot
[ -0.03882170468568802, -0.01603779010474682, -0.02078745886683464, 0.04307942092418671, 0.01999453268945217, 0.026762880384922028, 0.03698517009615898, 0.009271356277167797, -0.028615150600671768, -0.04553605616092682, -0.013976548798382282, 0.004361631814390421, -0.04939861223101616, 0.010...
WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents
https://openreview.net/forum?id=R9KnuFlvnU
[ "Shunyu Yao", "Howard Chen", "John Yang", "Karthik R Narasimhan" ]
null
null
Most existing benchmarks for grounding language in interactive environments either lack realistic linguistic elements, or prove difficult to scale up due to substantial human involvement in the collection of data or feedback signals. We develop WebShop – a simulated e-commerce website environment with 1.18 million real...
[ "language grounding", "reinforcement learning", "imitation learning", "natural language processing", "sim-to-real transfer", "web tasks" ]
We introduce a simulated website environment with real-world data to develop and test interactive agents with grounded language understanding.
12,341
2207.01206
title_snapshot
[ -0.034968338906764984, -0.011791682802140713, 0.0023028431460261345, 0.05215386673808098, 0.028100118041038513, 0.004019687883555889, 0.02461356669664383, 0.039506468921899796, 0.02363605797290802, -0.009039048105478287, -0.04898719862103462, 0.04601355642080307, -0.07692275941371918, -0.0...
On Infinite Separations Between Simple and Optimal Mechanisms
https://openreview.net/forum?id=Od4oKKwBx7Z
[ "Alexandros Psomas", "Ariel Schvartzman", "S. Matthew Weinberg" ]
null
null
We consider a revenue-maximizing seller with $k$ heterogeneous items for sale to a single additive buyer, whose values are drawn from a known, possibly correlated prior $\mathcal{D}$. It is known that there exist priors $\mathcal{D}$ such that simple mechanisms --- those with bounded menu complexity --- extract an arbi...
[ "mechanism design", "revenue maximization", "correlated distributions" ]
We characterize the structure of (correlated) distributions that witness an infinite separation between simple and optimal mechanisms.
12,334
2205.13039
title_snapshot
[ -0.03290664777159691, 0.002069941721856594, -0.00043805097811855376, 0.01973661035299301, 0.03633195534348488, 0.023101812228560448, 0.01647067628800869, -0.022275859490036964, -0.026185940951108932, -0.02595055289566517, 0.008441897109150887, 0.0062810503877699375, -0.05750901624560356, 0...
IMED-RL: Regret optimal learning of ergodic Markov decision processes
https://openreview.net/forum?id=kjR8GiwqCK
[ "Fabien Pesquerel", "Odalric-Ambrym Maillard" ]
null
null
We consider reinforcement learning in a discrete, undiscounted, infinite-horizon Markov decision problem (MDP) under the average reward criterion, and focus on the minimization of the regret with respect to an optimal policy, when the learner does not know the rewards nor transitions of the MDP. In light of their succ...
[ "Reinforcement Learning", "sequential learning", "average-reward", "lower bound", "regret minimization", "upper bound", "optimal", "Bandit", "ergodic" ]
We propose IMED-RL, a learning policy that is asymptotically optimal with respect to regret minimization problem under the average-reward criterion in ergodic MDPs with unknown reward and transition.
12,333
null
null
[ -0.0471639558672905, -0.012507128529250622, -0.02445693127810955, 0.043161675333976746, 0.052344586700201035, 0.0212969072163105, 0.025701967999339104, 0.009676029905676842, -0.03664734959602356, -0.0426100455224514, -0.0249472763389349, 0.00873144157230854, -0.054957494139671326, -0.04353...
Semantic Probabilistic Layers for Neuro-Symbolic Learning
https://openreview.net/forum?id=o-mxIWAY1T8
[ "Kareem Ahmed", "Stefano Teso", "Kai-Wei Chang", "Guy Van den Broeck", "Antonio Vergari" ]
null
null
We design a predictive layer for structured-output prediction (SOP) that can be plugged into any neural network guaranteeing its predictions are consistent with a set of predefined symbolic constraints. Our Semantic Probabilistic Layer (SPL) can model intricate correlations, and hard constraints, over a structured outp...
[ "neuro-symbolic learning", "structured-output prediction", "structured-output spaces", "tractable probabilistic inference", "probabilistic reasoning" ]
null
12,327
2206.00426
title_snapshot
[ -0.01990828663110733, -0.0096203638240695, -0.00898425467312336, 0.03303099796175957, 0.040503598749637604, 0.02491435408592224, 0.020670847967267036, -0.016454851254820824, -0.031036805361509323, -0.01035090908408165, -0.02443840727210045, 0.0166770052164793, -0.05880372226238251, -0.0167...