title
stringlengths
14
161
paper_url
stringlengths
45
61
authors
listlengths
1
28
type
stringclasses
0 values
primary_area
stringclasses
0 values
abstract
large_stringlengths
413
4.43k
keywords
listlengths
0
0
TL;DR
large_stringclasses
0 values
submission_number
int64
1
1.23k
arxiv_id
stringlengths
10
10
arxiv_id_source
stringclasses
2 values
embedding
listlengths
768
768
PAC-Bayesian Bounds on Rate-Efficient Classifiers
https://proceedings.mlr.press/v162/abbas22a.html
[ "Alhabib Abbas", "Yiannis Andreopoulos" ]
null
null
We derive analytic bounds on the noise invariance of majority vote classifiers operating on compressed inputs. Specifically, starting from recent bounds on the true risk of majority vote classifiers, we extend the applicability of PAC-Bayesian theory to quantify the resilience of majority votes to input noise stemming ...
[]
null
1
null
null
[ -0.021620197221636772, -0.0155895771458745, -0.04163597896695137, 0.05635705962777138, 0.025485912337899208, 0.03824486956000328, 0.028196390718221664, -0.03849766030907631, -0.032340649515390396, -0.028499407693743706, -0.020517397671937943, -0.016895458102226257, -0.07360758632421494, -0...
Sharp-MAML: Sharpness-Aware Model-Agnostic Meta Learning
https://proceedings.mlr.press/v162/abbas22b.html
[ "Momin Abbas", "Quan Xiao", "Lisha Chen", "Pin-Yu Chen", "Tianyi Chen" ]
null
null
Model-agnostic meta learning (MAML) is currently one of the dominating approaches for few-shot meta-learning. Albeit its effectiveness, the optimization of MAML can be challenging due to the innate bilevel problem structure. Specifically, the loss landscape of MAML is much more complex with possibly more saddle points...
[]
null
2
2206.03996
title_snapshot
[ -0.024057354778051376, -0.01719362661242485, 0.008802439086139202, 0.028284285217523575, 0.022634882479906082, 0.03779992088675499, 0.028298908844590187, -0.020523276180028915, -0.05628325790166855, 0.008209611289203167, -0.0027250386774539948, 0.0437278188765049, -0.06996995955705643, 0.0...
An Initial Alignment between Neural Network and Target is Needed for Gradient Descent to Learn
https://proceedings.mlr.press/v162/abbe22a.html
[ "Emmanuel Abbe", "Elisabetta Cornacchia", "Jan Hazla", "Christopher Marquis" ]
null
null
This paper introduces the notion of “Initial Alignment” (INAL) between a neural network at initialization and a target function. It is proved that if a network and a Boolean target function do not have a noticeable INAL, then noisy gradient descent with normalized i.i.d. initialization will not learn in polynomial time...
[]
null
3
2202.12846
title_snapshot
[ -0.02333051897585392, 0.006675853859633207, -0.029614703729748726, 0.013769001699984074, 0.03217779099941254, 0.027715807780623436, 0.038678739219903946, -0.008532569743692875, -0.03163059428334236, -0.018134871497750282, 0.001382990158163011, 0.022135350853204727, -0.05764763802289963, -0...
Active Sampling for Min-Max Fairness
https://proceedings.mlr.press/v162/abernethy22a.html
[ "Jacob D Abernethy", "Pranjal Awasthi", "Matthäus Kleindessner", "Jamie Morgenstern", "Chris Russell", "Jie Zhang" ]
null
null
We propose simple active sampling and reweighting strategies for optimizing min-max fairness that can be applied to any classification or regression model learned via loss minimization. The key intuition behind our approach is to use at each timestep a datapoint from the group that is worst off under the current model ...
[]
null
4
2006.06879
title_snapshot
[ -0.008334285579621792, -0.034243784844875336, -0.005701664835214615, 0.038151323795318604, 0.014649219810962677, 0.03954951837658882, -0.0072580622509121895, -0.010599506087601185, -0.03619711473584175, -0.059578198939561844, -0.011541463434696198, -0.00932457484304905, -0.08856764435768127,...
Meaningfully debugging model mistakes using conceptual counterfactual explanations
https://proceedings.mlr.press/v162/abid22a.html
[ "Abubakar Abid", "Mert Yuksekgonul", "James Zou" ]
null
null
Understanding and explaining the mistakes made by trained models is critical to many machine learning objectives, such as improving robustness, addressing concept drift, and mitigating biases. However, this is often an ad hoc process that involves manually looking at the model’s mistakes on many test samples and guessi...
[]
null
5
2106.12723
title_snapshot
[ -0.015277115628123283, -0.00947808101773262, -0.06517820060253143, 0.03409845381975174, 0.06768903136253357, 0.004244143143296242, 0.043111782521009445, 0.006185307167470455, -0.008772323839366436, -0.05477622523903847, -0.011568153277039528, 0.045281387865543365, -0.04794757068157196, 0.0...
Batched Dueling Bandits
https://proceedings.mlr.press/v162/agarwal22a.html
[ "Arpit Agarwal", "Rohan Ghuge", "Viswanath Nagarajan" ]
null
null
The K-armed dueling bandit problem, where the feedback is in the form of noisy pairwise comparisons, has been widely studied. Previous works have only focused on the sequential setting where the policy adapts after every comparison. However, in many applications such as search ranking and recommendation systems, it is ...
[]
null
6
2202.10660
title_snapshot
[ -0.023853644728660583, -0.014535333029925823, -0.010070433840155602, 0.03752369061112404, 0.00008459947275696322, 0.0286223366856575, 0.029062043875455856, 0.016932664439082146, -0.024646254256367683, -0.04846819117665291, -0.004670827183872461, 0.01026197336614132, -0.05993778258562088, -...
Hierarchical Shrinkage: Improving the accuracy and interpretability of tree-based models.
https://proceedings.mlr.press/v162/agarwal22b.html
[ "Abhineet Agarwal", "Yan Shuo Tan", "Omer Ronen", "Chandan Singh", "Bin Yu" ]
null
null
Decision trees and random forests (RF) are a cornerstone of modern machine learning practice. Due to their tendency to overfit, trees are typically regularized by a variety of techniques that modify their structure (e.g. pruning). We introduce Hierarchical Shrinkage (HS), a post-hoc algorithm which regularizes the tree...
[]
null
7
2202.00858
title_judge
[ -0.012564145028591156, 0.0020067167934030294, -0.0017450093291699886, 0.029640132561326027, 0.04612687602639198, 0.04138323292136192, 0.048238370567560196, -0.0383136160671711, -0.03786083683371544, -0.04855704307556152, -0.01958075352013111, -0.00458461744710803, -0.07202678173780441, 0.0...
Deep equilibrium networks are sensitive to initialization statistics
https://proceedings.mlr.press/v162/agarwala22a.html
[ "Atish Agarwala", "Samuel S Schoenholz" ]
null
null
Deep equilibrium networks (DEQs) are a promising way to construct models which trade off memory for compute. However, theoretical understanding of these models is still lacking compared to traditional networks, in part because of the repeated application of a single set of weights. We show that DEQs are sensitive to th...
[]
null
8
2207.09432
title_snapshot
[ -0.02127249725162983, -0.012307806871831417, -0.008009486831724644, 0.03484112024307251, 0.03499886021018028, 0.032464850693941116, -0.002058434998616576, 0.003875507740303874, -0.011562538333237171, -0.03725374862551689, 0.01258148904889822, 0.012122565880417824, -0.07518792152404785, -0....
Learning of Cluster-based Feature Importance for Electronic Health Record Time-series
https://proceedings.mlr.press/v162/aguiar22a.html
[ "Henrique Aguiar", "Mauro Santos", "Peter Watkinson", "Tingting Zhu" ]
null
null
The recent availability of Electronic Health Records (EHR) has allowed for the development of algorithms predicting inpatient risk of deterioration and trajectory evolution. However, prediction of disease progression with EHR is challenging since these data are sparse, heterogeneous, multi-dimensional, and multi-modal ...
[]
null
9
null
null
[ -0.01694558374583721, -0.015721777454018593, -0.0008803087985143065, 0.013484636321663857, 0.040025003254413605, 0.03591537848114967, 0.03206395357847214, -0.0032681766897439957, -0.030780183151364326, -0.03559483587741852, -0.0006113044219091535, -0.005797790363430977, -0.04435497894883156,...
On the Convergence of the Shapley Value in Parametric Bayesian Learning Games
https://proceedings.mlr.press/v162/agussurja22a.html
[ "Lucas Agussurja", "Xinyi Xu", "Bryan Kian Hsiang Low" ]
null
null
Measuring contributions is a classical problem in cooperative game theory where the Shapley value is the most well-known solution concept. In this paper, we establish the convergence property of the Shapley value in parametric Bayesian learning games where players perform a Bayesian inference using their combined data,...
[]
null
10
2205.07428
title_snapshot
[ -0.045522239059209824, -0.00945024099200964, 0.013294278644025326, 0.02758736163377762, 0.03355932608246803, 0.023665975779294968, 0.014704329892992973, 0.0012164918007329106, -0.018482660874724388, -0.04335298389196396, 0.002070111222565174, 0.00880051776766777, -0.06082694232463837, 0.00...
Individual Preference Stability for Clustering
https://proceedings.mlr.press/v162/ahmadi22a.html
[ "Saba Ahmadi", "Pranjal Awasthi", "Samir Khuller", "Matthäus Kleindessner", "Jamie Morgenstern", "Pattara Sukprasert", "Ali Vakilian" ]
null
null
In this paper, we propose a natural notion of individual preference (IP) stability for clustering, which asks that every data point, on average, is closer to the points in its own cluster than to the points in any other cluster. Our notion can be motivated from several perspectives, including game theory and algorithmi...
[]
null
11
2207.03600
title_snapshot
[ -0.02298051118850708, -0.008229962550103664, 0.02073863334953785, 0.045761119574308395, 0.047764502465724945, 0.0336461216211319, 0.0044381204061210155, 0.00257417862303555, -0.03282289579510689, -0.05027539283037186, -0.020579801872372627, -0.028004666790366173, -0.07300779968500137, -0.0...
Understanding the unstable convergence of gradient descent
https://proceedings.mlr.press/v162/ahn22a.html
[ "Kwangjun Ahn", "Jingzhao Zhang", "Suvrit Sra" ]
null
null
Most existing analyses of (stochastic) gradient descent rely on the condition that for $L$-smooth costs, the step size is less than $2/L$. However, many works have observed that in machine learning applications step sizes often do not fulfill this condition, yet (stochastic) gradient descent still converges, albeit in ...
[]
null
12
2204.01050
title_snapshot
[ -0.046942081302404404, -0.027417749166488647, 0.01007110346108675, 0.034856848418712616, 0.0508061982691288, 0.013734686188399792, 0.021657440811395645, -0.00013348858919925988, -0.010436700657010078, -0.048850275576114655, -0.015779489651322365, -0.005627320613712072, -0.0622515007853508, ...
Minimum Cost Intervention Design for Causal Effect Identification
https://proceedings.mlr.press/v162/akbari22a.html
[ "Sina Akbari", "Jalal Etesami", "Negar Kiyavash" ]
null
null
Pearl’s do calculus is a complete axiomatic approach to learn the identifiable causal effects from observational data. When such an effect is not identifiable, it is necessary to perform a collection of often costly interventions in the system to learn the causal effect. In this work, we consider the problem of designi...
[]
null
13
null
null
[ -0.011175736784934998, 0.0028737932443618774, -0.027302183210849762, 0.033806949853897095, 0.054501380771398544, 0.03224358335137367, 0.031077060848474503, -0.0012402931461110711, 0.013487015850841999, -0.021074162796139717, 0.000670753710437566, 0.04474056512117386, -0.052142184227705, -0...
How Faithful is your Synthetic Data? Sample-level Metrics for Evaluating and Auditing Generative Models
https://proceedings.mlr.press/v162/alaa22a.html
[ "Ahmed Alaa", "Boris Van Breugel", "Evgeny S. Saveliev", "Mihaela van der Schaar" ]
null
null
Devising domain- and model-agnostic evaluation metrics for generative models is an important and as yet unresolved problem. Most existing metrics, which were tailored solely to the image synthesis setup, exhibit a limited capacity for diagnosing the different modes of failure of generative models across broader applica...
[]
null
14
2102.08921
title_snapshot
[ 0.0017761958297342062, -0.02101946622133255, -0.00990328285843134, 0.06294669210910797, 0.04981759563088417, 0.008481848984956741, 0.03780767321586609, -0.014733078889548779, -0.008505214937031269, -0.0461963452398777, -0.017115026712417603, -0.003401579800993204, -0.0791371762752533, 0.00...
A Natural Actor-Critic Framework for Zero-Sum Markov Games
https://proceedings.mlr.press/v162/alacaoglu22a.html
[ "Ahmet Alacaoglu", "Luca Viano", "Niao He", "Volkan Cevher" ]
null
null
We introduce algorithms based on natural actor-critic and analyze their sample complexity for solving two player zero-sum Markov games in the tabular case. Our results improve the best-known sample complexities of policy gradient/actor-critic methods for convergence to Nash equilibrium in the multi-agent setting. We us...
[]
null
15
null
null
[ -0.04779203236103058, -0.027632329612970352, 0.006226781290024519, 0.04454214870929718, 0.05165131390094757, 0.027805017307400703, 0.020209666341543198, 0.01711348071694374, -0.03760310262441635, -0.031052421778440475, -0.005234254524111748, 0.04509748890995979, -0.08414416760206223, -0.01...
Deploying Convolutional Networks on Untrusted Platforms Using 2D Holographic Reduced Representations
https://proceedings.mlr.press/v162/alam22a.html
[ "Mohammad Mahmudul Alam", "Edward Raff", "Tim Oates", "James Holt" ]
null
null
Due to the computational cost of running inference for a neural network, the need to deploy the inferential steps on a third party’s compute environment or hardware is common. If the third party is not fully trusted, it is desirable to obfuscate the nature of the inputs and outputs, so that the third party can not easi...
[]
null
16
2206.05893
title_snapshot
[ -0.006858104839920998, 0.008400068618357182, -0.02719334326684475, 0.0818747878074646, 0.04291554540395737, -0.004784626420587301, 0.008179433643817902, -0.004014736972749233, -0.02048465609550476, -0.00694150198251009, -0.008625953458249569, -0.01247359998524189, -0.06361175328493118, 0.0...
Optimistic Linear Support and Successor Features as a Basis for Optimal Policy Transfer
https://proceedings.mlr.press/v162/alegre22a.html
[ "Lucas Nunes Alegre", "Ana Bazzan", "Bruno C. Da Silva" ]
null
null
In many real-world applications, reinforcement learning (RL) agents might have to solve multiple tasks, each one typically modeled via a reward function. If reward functions are expressed linearly, and the agent has previously learned a set of policies for different tasks, successor features (SFs) can be exploited to c...
[]
null
17
2206.11326
title_snapshot
[ -0.0328923836350441, -0.036725178360939026, 0.02838248945772648, 0.04831196740269661, 0.044536564499139786, 0.031157387420535088, -0.006908811163157225, -0.01678028330206871, -0.027629492804408073, -0.042649045586586, -0.009890222921967506, 0.02122923545539379, -0.09629429876804352, -0.024...
Structured Stochastic Gradient MCMC
https://proceedings.mlr.press/v162/alexos22a.html
[ "Antonios Alexos", "Alex J Boyd", "Stephan Mandt" ]
null
null
Stochastic gradient Markov Chain Monte Carlo (SGMCMC) is a scalable algorithm for asymptotically exact Bayesian inference in parameter-rich models, such as Bayesian neural networks. However, since mixing can be slow in high dimensions, practitioners often resort to variational inference (VI). Unfortunately, VI makes st...
[]
null
18
2107.09028
title_snapshot
[ -0.006625790148973465, -0.010323278605937958, -0.016504894942045212, 0.04686397686600685, 0.0369848757982254, 0.03982171788811684, 0.024427475407719612, -0.0030048643238842487, -0.029830971732735634, -0.03969563543796539, 0.017525063827633858, 0.005794597323983908, -0.04812713712453842, 0....
XAI for Transformers: Better Explanations through Conservative Propagation
https://proceedings.mlr.press/v162/ali22a.html
[ "Ameen Ali", "Thomas Schnake", "Oliver Eberle", "Grégoire Montavon", "Klaus-Robert Müller", "Lior Wolf" ]
null
null
Transformers have become an important workhorse of machine learning, with numerous applications. This necessitates the development of reliable methods for increasing their transparency. Multiple interpretability methods, often based on gradient information, have been proposed. We show that the gradient in a Transformer...
[]
null
19
2202.07304
title_snapshot
[ -0.012018230743706226, -0.03011173941195011, -0.013728911057114601, 0.04022175446152687, 0.03128911927342415, 0.03211405500769615, 0.00011836408521048725, -0.0021589654497802258, -0.010680371895432472, -0.015801917761564255, -0.05104782059788704, 0.04574302211403847, -0.046410564333200455, ...
RUMs from Head-to-Head Contests
https://proceedings.mlr.press/v162/almanza22a.html
[ "Matteo Almanza", "Flavio Chierichetti", "Ravi Kumar", "Alessandro Panconesi", "Andrew Tomkins" ]
null
null
Random utility models (RUMs) encode the likelihood that a particular item will be selected from a slate of competing items. RUMs are well-studied objects in both discrete choice theory and, more recently, in the machine learning community, as they encode a fairly broad notion of rational user behavior. In this paper, w...
[]
null
20
null
null
[ -0.0410955436527729, -0.006871474906802177, -0.01780509017407894, 0.031534500420093536, 0.013589196838438511, 0.021334465593099594, 0.007500501815229654, 0.04130076617002487, -0.022130770608782768, -0.03675350546836853, -0.018188530579209328, 0.0071299937553703785, -0.05836563557386398, -0...
Neuro-Symbolic Language Modeling with Automaton-augmented Retrieval
https://proceedings.mlr.press/v162/alon22a.html
[ "Uri Alon", "Frank Xu", "Junxian He", "Sudipta Sengupta", "Dan Roth", "Graham Neubig" ]
null
null
Retrieval-based language models (R-LM) model the probability of natural language text by combining a standard language model (LM) with examples retrieved from an external datastore at test time. While effective, a major bottleneck of using these models in practice is the computationally costly datastore search, which c...
[]
null
21
2201.12431
title_snapshot
[ -0.0419473722577095, -0.022128066048026085, -0.03334774821996689, 0.03782501071691513, 0.05156364291906357, 0.05423949286341667, 0.032200999557971954, 0.014642667956650257, -0.0327426940202713, 0.00007028758409433067, -0.032328300178050995, 0.03746918588876724, -0.059357162564992905, -0.02...
Minimax Classification under Concept Drift with Multidimensional Adaptation and Performance Guarantees
https://proceedings.mlr.press/v162/alvarez22a.html
[ "Verónica Álvarez", "Santiago Mazuelas", "Jose A Lozano" ]
null
null
The statistical characteristics of instance-label pairs often change with time in practical scenarios of supervised classification. Conventional learning techniques adapt to such concept drift accounting for a scalar rate of change by means of a carefully chosen learning rate, forgetting factor, or window size. However...
[]
null
22
2205.15942
title_snapshot
[ -0.013423060066998005, -0.0336882658302784, -0.013898244127631187, 0.03670521080493927, 0.044716976583004, 0.027611901983618736, 0.027546577155590057, -0.01738211326301098, -0.01638575829565525, -0.05254686623811722, -0.023000696673989296, 0.017460225149989128, -0.04675184190273285, 0.0268...
Scalable First-Order Bayesian Optimization via Structured Automatic Differentiation
https://proceedings.mlr.press/v162/ament22a.html
[ "Sebastian E Ament", "Carla P Gomes" ]
null
null
Bayesian Optimization (BO) has shown great promise for the global optimization of functions that are expensive to evaluate, but despite many successes, standard approaches can struggle in high dimensions. To improve the performance of BO, prior work suggested incorporating gradient information into a Gaussian process s...
[]
null
23
2206.08366
title_snapshot
[ -0.0309464018791914, -0.007947216741740704, 0.0328887514770031, 0.03181120753288269, 0.021469347178936005, 0.04714420810341835, 0.015144423581659794, -0.04635993763804436, 0.0012158420868217945, -0.0324159599840641, 0.0034012822434306145, 0.017133397981524467, -0.055792562663555145, -0.001...
Public Data-Assisted Mirror Descent for Private Model Training
https://proceedings.mlr.press/v162/amid22a.html
[ "Ehsan Amid", "Arun Ganesh", "Rajiv Mathews", "Swaroop Ramaswamy", "Shuang Song", "Thomas Steinke", "Thomas Steinke", "Vinith M Suriyakumar", "Om Thakkar", "Abhradeep Thakurta" ]
null
null
In this paper, we revisit the problem of using in-distribution public data to improve the privacy/utility trade-offs for differentially private (DP) model training. (Here, public data refers to auxiliary data sets that have no privacy concerns.) We design a natural variant of DP mirror descent, where the DP gradients o...
[]
null
24
2112.00193
title_snapshot
[ -0.018890568986535072, -0.011607310734689236, -0.001918046735227108, 0.07440412789583206, 0.03297426179051399, 0.033724475651979446, 0.0571100227534771, -0.037487443536520004, -0.0023606636095792055, -0.02510642446577549, -0.0013741268776357174, 0.006920310202986002, -0.07275386154651642, ...
On Last-Iterate Convergence Beyond Zero-Sum Games
https://proceedings.mlr.press/v162/anagnostides22a.html
[ "Ioannis Anagnostides", "Ioannis Panageas", "Gabriele Farina", "Tuomas Sandholm" ]
null
null
Most existing results about last-iterate convergence of learning dynamics are limited to two-player zero-sum games, and only apply under rigid assumptions about what dynamics the players follow. In this paper we provide new results and techniques that apply to broader families of games and learning dynamics. First, we ...
[]
null
25
2203.12056
title_snapshot
[ -0.06273333728313446, -0.023810558021068573, 0.04456254839897156, 0.027626236900687218, 0.028760217130184174, 0.025826070457696915, 0.012651235796511173, 0.04337429255247116, -0.03432151675224304, -0.026191016659140587, 0.0064800940454006195, 0.010820716619491577, -0.06156741827726364, -0....
Online Algorithms with Multiple Predictions
https://proceedings.mlr.press/v162/anand22a.html
[ "Keerti Anand", "Rong Ge", "Amit Kumar", "Debmalya Panigrahi" ]
null
null
This paper studies online algorithms augmented withmultiplemachine-learned predictions. We give a generic algorithmic framework for online covering problems with multiple predictions that obtains an online solution that is competitive against the performance of thebestsolution obtained from the predictions. Our algorit...
[]
null
26
2205.03921
title_snapshot
[ -0.013217519968748093, -0.007974378764629364, 0.007209242321550846, 0.031080609187483788, 0.05315132811665535, 0.04526729881763458, 0.004309363663196564, 0.018232598900794983, -0.034766558557748795, -0.0285917017608881, -0.005690871737897396, -0.005344747100025415, -0.08555108308792114, -0...
Learning to Hash Robustly, Guaranteed
https://proceedings.mlr.press/v162/andoni22a.html
[ "Alexandr Andoni", "Daniel Beaglehole" ]
null
null
The indexing algorithms for the high-dimensional nearest neighbor search (NNS) with the best worst-case guarantees are based on the randomized Locality Sensitive Hashing (LSH), and its derivatives. In practice, many heuristic approaches exist to "learn" the best indexing method in order to speed-up NNS, crucially adapt...
[]
null
27
2108.05433
title_snapshot
[ -0.03616471216082573, -0.016223832964897156, -0.00866163894534111, 0.057234227657318115, 0.03744989261031151, 0.020753268152475357, 0.01646295003592968, -0.014812787994742393, -0.016086356714367867, -0.0335364006459713, -0.0118646789342165, -0.032797928899526596, -0.04578876495361328, 0.00...
Set Based Stochastic Subsampling
https://proceedings.mlr.press/v162/andreis22a.html
[ "Bruno Andreis", "Seanie Lee", "A. Tuan Nguyen", "Juho Lee", "Eunho Yang", "Sung Ju Hwang" ]
null
null
Deep models are designed to operate on huge volumes of high dimensional data such as images. In order to reduce the volume of data these models must process, we propose a set-based two-stage end-to-end neural subsampling model that is jointly optimized with anarbitrarydownstream task network (e.g. classifier). In the f...
[]
null
28
2006.14222
title_snapshot
[ 0.0286420788615942, -0.0266202911734581, -0.021119650453329086, 0.05149325355887413, 0.046726372092962265, 0.044550999999046326, 0.0033071052748709917, -0.0014788996195420623, -0.016277870163321495, -0.06320080161094666, -0.018864259123802185, 0.00006722776015521958, -0.05941391363739967, ...
Towards Understanding Sharpness-Aware Minimization
https://proceedings.mlr.press/v162/andriushchenko22a.html
[ "Maksym Andriushchenko", "Nicolas Flammarion" ]
null
null
Sharpness-Aware Minimization (SAM) is a recent training method that relies on worst-case weight perturbations which significantly improves generalization in various settings. We argue that the existing justifications for the success of SAM which are based on a PAC-Bayes generalization bound and the idea of convergence ...
[]
null
29
2206.06232
title_snapshot
[ -0.028070524334907532, -0.021225320175290108, 0.02783988229930401, 0.05071419104933739, 0.01636086031794548, 0.04888499155640602, 0.039827000349760056, 0.0075206635519862175, -0.036414436995983124, -0.050221092998981476, -0.025824489071965218, 0.01865503005683422, -0.06562428921461105, 0.0...
Fair and Fast k-Center Clustering for Data Summarization
https://proceedings.mlr.press/v162/angelidakis22a.html
[ "Haris Angelidakis", "Adam Kurpisz", "Leon Sering", "Rico Zenklusen" ]
null
null
We consider two key issues faced by many clustering methods when used for data summarization, namely (a) an unfair representation of "demographic groups” and (b) distorted summarizations, where data points in the summary represent subsets of the original data of vastly different sizes. Previous work made important step...
[]
null
30
null
null
[ -0.03928057476878166, -0.04084715619683266, -0.007531350944191217, 0.030758438631892204, 0.034942105412483215, 0.03071572631597519, 0.005050340201705694, 0.010385571047663689, -0.030848698690533638, -0.010428007692098618, -0.01830105483531952, -0.020791126415133476, -0.06947945058345795, 0...
Interactive Correlation Clustering with Existential Cluster Constraints
https://proceedings.mlr.press/v162/angell22a.html
[ "Rico Angell", "Nicholas Monath", "Nishant Yadav", "Andrew Mccallum" ]
null
null
We consider the problem of clustering with user feedback. Existing methods express constraints about the input data points, most commonly through must-link and cannot-link constraints on data point pairs. In this paper, we introduce existential cluster constraints: a new form of feedback where users indicate the featur...
[]
null
31
null
null
[ 0.005291120149195194, 0.0029497083742171526, 0.002218895824626088, 0.0465603843331337, 0.053727880120277405, 0.0349702462553978, 0.010459153912961483, -0.00026749534299597144, -0.03187236189842224, -0.019946854561567307, -0.049275245517492294, -0.00648296345025301, -0.05585533007979393, -0...
Image-to-Image Regression with Distribution-Free Uncertainty Quantification and Applications in Imaging
https://proceedings.mlr.press/v162/angelopoulos22a.html
[ "Anastasios N Angelopoulos", "Amit Pal Kohli", "Stephen Bates", "Michael Jordan", "Jitendra Malik", "Thayer Alshaabi", "Srigokul Upadhyayula", "Yaniv Romano" ]
null
null
Image-to-image regression is an important learning task, used frequently in biological imaging. Current algorithms, however, do not generally offer statistical guarantees that protect against a model’s mistakes and hallucinations. To address this, we develop uncertainty quantification techniques with rigorous statistic...
[]
null
32
2202.05265
title_snapshot
[ 0.004752897657454014, 0.00514919264242053, -0.041496023535728455, 0.04306650906801224, 0.0510389469563961, 0.030839184299111366, 0.02528962306678295, -0.01384506281465292, -0.03513864427804947, -0.04969542846083641, -0.02595304325222969, 0.0097417663782835, -0.06273192912340164, 0.03243326...
AdaGrad Avoids Saddle Points
https://proceedings.mlr.press/v162/antonakopoulos22a.html
[ "Kimon Antonakopoulos", "Panayotis Mertikopoulos", "Georgios Piliouras", "Xiao Wang" ]
null
null
Adaptive first-order methods in optimization have widespread ML applications due to their ability to adapt to non-convex landscapes. However, their convergence guarantees are typically stated in terms of vanishing gradient norms, which leaves open the issue of converging to undesirable saddle points (or even local maxi...
[]
null
33
null
null
[ -0.061166949570178986, -0.025921786203980446, 0.009317640215158463, 0.034699760377407074, 0.04057425633072853, 0.02006986364722252, 0.056762684136629105, -0.030684305354952812, -0.0265949834138155, -0.06124249845743179, 0.001390657969750464, -0.01082840096205473, -0.053084034472703934, 0.0...
UnderGrad: A Universal Black-Box Optimization Method with Almost Dimension-Free Convergence Rate Guarantees
https://proceedings.mlr.press/v162/antonakopoulos22b.html
[ "Kimon Antonakopoulos", "Dong Quan Vu", "Volkan Cevher", "Kfir Levy", "Panayotis Mertikopoulos" ]
null
null
Universal methods achieve optimal convergence rate guarantees in convex optimization without any prior knowledge of the problem’s regularity parameters or the attributes of the gradient oracle employed by the method. In this regard, existing state-of-the-art algorithms achieve an $O(1/T^2)$ convergence rate in Lipschit...
[]
null
34
2206.09352
title_judge
[ -0.036589041352272034, -0.013027826324105263, 0.028969144448637962, 0.030559366568922997, 0.031345538794994354, 0.05818968266248703, 0.021921133622527122, 0.026600247249007225, -0.014385383576154709, -0.03152000904083252, -0.01939132995903492, -0.008887384086847305, -0.04521329700946808, -...
Adapting the Linearised Laplace Model Evidence for Modern Deep Learning
https://proceedings.mlr.press/v162/antoran22a.html
[ "Javier Antoran", "David Janz", "James U Allingham", "Erik Daxberger", "Riccardo Rb Barbano", "Eric Nalisnick", "Jose Miguel Hernandez-Lobato" ]
null
null
The linearised Laplace method for estimating model uncertainty has received renewed attention in the Bayesian deep learning community. The method provides reliable error bars and admits a closed-form expression for the model evidence, allowing for scalable selection of model hyperparameters. In this work, we examine th...
[]
null
35
2206.08900
title_snapshot
[ 0.0005298020550981164, -0.030669325962662697, -0.009981878101825714, 0.009477611631155014, 0.043577536940574646, 0.05563076585531235, 0.02665753662586212, -0.01983478292822838, -0.013638313859701157, -0.06350405514240265, 0.004396350122988224, -0.010482823476195335, -0.07150959223508835, 0...
EAT-C: Environment-Adversarial sub-Task Curriculum for Efficient Reinforcement Learning
https://proceedings.mlr.press/v162/ao22a.html
[ "Shuang Ao", "Tianyi Zhou", "Jing Jiang", "Guodong Long", "Xuan Song", "Chengqi Zhang" ]
null
null
Reinforcement learning (RL) is inefficient on long-horizon tasks due to sparse rewards and its policy can be fragile to slightly perturbed environments. We address these challenges via a curriculum of tasks with coupled environments, generated by two policies trained jointly with RL: (1) a co-operative planning policy ...
[]
null
36
null
null
[ -0.002013680525124073, -0.033110372722148895, -0.013044280000030994, 0.0507061704993248, 0.05920088291168213, 0.007380381692200899, 0.02450125478208065, -0.004019662272185087, -0.03189396485686302, -0.030084678903222084, -0.05377476289868355, 0.03519074618816376, -0.06788934022188187, -0.0...
Online Balanced Experimental Design
https://proceedings.mlr.press/v162/arbour22a.html
[ "David Arbour", "Drew Dimmery", "Tung Mai", "Anup Rao" ]
null
null
We consider the experimental design problem in an online environment, an important practical task for reducing the variance of estimates in randomized experiments which allows for greater precision, and in turn, improved decision making. In this work, we present algorithms that build on recent advances in online discre...
[]
null
37
2203.02025
title_snapshot
[ 0.002825577510520816, 0.0006557967863045633, -0.024329420179128647, 0.029354672878980637, 0.04247157648205757, 0.038037389516830444, 0.02750135026872158, -0.0015496459091082215, -0.002547991229221225, -0.05178007856011391, 0.004995929542928934, -0.004596857354044914, -0.07599461078643799, ...
VariGrow: Variational Architecture Growing for Task-Agnostic Continual Learning based on Bayesian Novelty
https://proceedings.mlr.press/v162/ardywibowo22a.html
[ "Randy Ardywibowo", "Zepeng Huo", "Zhangyang Wang", "Bobak J Mortazavi", "Shuai Huang", "Xiaoning Qian" ]
null
null
Continual Learning (CL) is the problem of sequentially learning a set of tasks and preserving all the knowledge acquired. Many existing methods assume that the data stream is explicitly divided into a sequence of known contexts (tasks), and use this information to know when to transfer knowledge from one context to ano...
[]
null
38
null
null
[ 0.03242767974734306, -0.04246705025434494, 0.0037180031649768353, 0.011591842398047447, 0.03104989416897297, 0.011078969575464725, 0.04645908251404762, 0.008646663278341293, -0.04701671376824379, -0.06066923215985298, -0.023180240765213966, 0.017810313031077385, -0.06752156466245651, 0.018...
Thresholded Lasso Bandit
https://proceedings.mlr.press/v162/ariu22a.html
[ "Kaito Ariu", "Kenshi Abe", "Alexandre Proutiere" ]
null
null
In this paper, we revisit the regret minimization problem in sparse stochastic contextual linear bandits, where feature vectors may be of large dimension $d$, but where the reward function depends on a few, say $s_0\ll d$, of these features only. We present Thresholded Lasso bandit, an algorithm that (i) estimates the ...
[]
null
39
2010.11994
title_snapshot
[ -0.01570107974112034, -0.005502893589437008, 0.023417029529809952, 0.020890112966299057, 0.027761608362197876, 0.03691154718399048, 0.03601529449224472, 0.006102308165282011, -0.047047846019268036, -0.03903970494866371, -0.005385556723922491, 0.0020405889954417944, -0.04784806817770004, -0...
Gradient Based Clustering
https://proceedings.mlr.press/v162/armacki22a.html
[ "Aleksandar Armacki", "Dragana Bajovic", "Dusan Jakovetic", "Soummya Kar" ]
null
null
We propose a general approach for distance based clustering, using the gradient of the cost function that measures clustering quality with respect to cluster assignments and cluster center positions. The approach is an iterative two step procedure (alternating between cluster assignment and cluster center updates) and ...
[]
null
40
2202.00720
title_snapshot
[ -0.009464395232498646, -0.00014047540025785565, 0.0037387420888990164, 0.03167635202407837, 0.0403701476752758, 0.06445689499378204, -0.0006193302106112242, -0.00543230352923274, -0.017621010541915894, -0.04137812554836273, -0.013929047621786594, -0.006374802906066179, -0.06177995353937149, ...
Understanding Gradient Descent on the Edge of Stability in Deep Learning
https://proceedings.mlr.press/v162/arora22a.html
[ "Sanjeev Arora", "Zhiyuan Li", "Abhishek Panigrahi" ]
null
null
Deep learning experiments by \citet{cohen2021gradient} using deterministic Gradient Descent (GD) revealed anEdge of Stability (EoS)phase when learning rate (LR) and sharpness (i.e., the largest eigenvalue of Hessian) no longer behave as in traditional optimization. Sharpness stabilizes around $2/$LR and loss goes up an...
[]
null
41
2205.09745
title_judge
[ -0.03692435100674629, -0.02710568532347679, 0.012866517528891563, 0.029414767399430275, 0.04147036001086235, 0.02578909508883953, 0.03607090190052986, 0.0016059844056144357, -0.04643908143043518, -0.056576769798994064, -0.008205276913940907, -0.006116820964962244, -0.06133817136287689, 0.0...
Private optimization in the interpolation regime: faster rates and hardness results
https://proceedings.mlr.press/v162/asi22a.html
[ "Hilal Asi", "Karan Chadha", "Gary Cheng", "John Duchi" ]
null
null
In non-private stochastic convex optimization, stochastic gradient methods converge much faster on interpolation problems—namely, problems where there exists a solution that simultaneously minimizes all of the sample losses—than on non-interpolating ones; similar improvements are not known in the private setting. In th...
[]
null
42
2210.17070
title_snapshot
[ -0.01137457974255085, 0.018419411033391953, 0.016525907441973686, 0.05824990198016167, 0.036944713443517685, 0.0455184280872345, 0.039957720786333084, -0.017415456473827362, 0.0010998838115483522, -0.040185652673244476, -0.0019166129641234875, -0.012303122319281101, -0.03879064694046974, -...
Optimal Algorithms for Mean Estimation under Local Differential Privacy
https://proceedings.mlr.press/v162/asi22b.html
[ "Hilal Asi", "Vitaly Feldman", "Kunal Talwar" ]
null
null
We study the problem of mean estimation of $\ell_2$-bounded vectors under the constraint of local differential privacy. While the literature has a variety of algorithms that achieve the (asymptotic) optimal rates for this problem, the performance of these algorithms in practice can vary significantly due to varying (an...
[]
null
43
2205.02466
title_snapshot
[ -0.01628575474023819, 0.012506322003901005, 0.010565477423369884, 0.046466927975416183, 0.03670138493180275, 0.02463202178478241, 0.05580440163612366, -0.01471615955233574, -0.031239662319421768, -0.036061543971300125, 0.01729723811149597, -0.041590575128793716, -0.05844879150390625, -0.01...
Asymptotically-Optimal Gaussian Bandits with Side Observations
https://proceedings.mlr.press/v162/atsidakou22a.html
[ "Alexia Atsidakou", "Orestis Papadigenopoulos", "Constantine Caramanis", "Sujay Sanghavi", "Sanjay Shakkottai" ]
null
null
We study the problem of Gaussian bandits with general side information, as first introduced by Wu, Szepesvári, and György. In this setting, the play of an arm reveals information about other arms, according to an arbitrarya prioriknownside informationmatrix: each element of this matrix encodes the fidelity of the infor...
[]
null
44
2505.10698
title_snapshot
[ -0.029092324897646904, -0.007167916279286146, 0.009730666875839233, 0.021755550056695938, 0.014549460262060165, 0.03234516829252243, 0.02729586325585842, 0.016791611909866333, -0.0150142852216959, -0.044527094811201096, 0.0126144178211689, 0.016645455732941628, -0.07168830931186676, -0.030...
Congested Bandits: Optimal Routing via Short-term Resets
https://proceedings.mlr.press/v162/awasthi22a.html
[ "Pranjal Awasthi", "Kush Bhatia", "Sreenivas Gollapudi", "Kostas Kollias" ]
null
null
For traffic routing platforms, the choice of which route to recommend to a user depends on the congestion on these routes – indeed, an individual’s utility depends on the number of people using the recommended route at that instance. Motivated by this, we introduce the problem of Congested Bandits where each arm’s rewa...
[]
null
45
2301.09251
title_snapshot
[ -0.012771001085639, -0.04170214757323265, 0.0032916096970438957, 0.03816402330994606, 0.027891213074326515, 0.025953996926546097, 0.026784365996718407, 0.06541844457387924, -0.04539410024881363, -0.05804234743118286, 0.0030172693077474833, -0.011342497542500496, -0.06271117180585861, -0.02...
Do More Negative Samples Necessarily Hurt In Contrastive Learning?
https://proceedings.mlr.press/v162/awasthi22b.html
[ "Pranjal Awasthi", "Nishanth Dikkala", "Pritish Kamath" ]
null
null
Recent investigations in noise contrastive estimation suggest, both empirically as well as theoretically, that while having more “negative samples” in the contrastive loss improves downstream classification performance initially, beyond a threshold, it hurts downstream performance due to a “collision-coverage” trade-of...
[]
null
46
2205.01789
title_snapshot
[ -0.007451306562870741, -0.011499118991196156, 0.005655348300933838, 0.05853917449712753, 0.009518404491245747, 0.01735898107290268, 0.016574677079916, -0.007435216568410397, -0.017109757289290428, -0.030650673434138298, 0.012564914301037788, 0.02234671264886856, -0.0754447877407074, -0.007...
H-Consistency Bounds for Surrogate Loss Minimizers
https://proceedings.mlr.press/v162/awasthi22c.html
[ "Pranjal Awasthi", "Anqi Mao", "Mehryar Mohri", "Yutao Zhong" ]
null
null
We present a detailed study of estimation errors in terms of surrogate loss estimation errors. We refer to such guarantees as H-consistency bounds, since they account for the hypothesis set H adopted. These guarantees are significantly stronger than H-calibration or H-consistency. They are also more informative than si...
[]
null
47
2205.08017
title_judge
[ -0.017781315371394157, 0.015097556635737419, -0.014981786720454693, 0.051617734134197235, 0.02576432004570961, 0.054315902292728424, 0.022832343354821205, -0.033289190381765366, -0.024558907374739647, -0.036948177963495255, -0.0004799425951205194, 0.009860508143901825, -0.06518808007240295, ...
Iterative Hard Thresholding with Adaptive Regularization: Sparser Solutions Without Sacrificing Runtime
https://proceedings.mlr.press/v162/axiotis22a.html
[ "Kyriakos Axiotis", "Maxim Sviridenko" ]
null
null
We propose a simple modification to the iterative hard thresholding (IHT) algorithm, which recovers asymptotically sparser solutions as a function of the condition number. When aiming to minimize a convex function f(x) with condition number $\kappa$ subject to x being an s-sparse vector, the standard IHT guarantee is a...
[]
null
48
2204.08274
title_snapshot
[ -0.024334214627742767, -0.003820119658485055, 0.04663082957267761, 0.033071357756853104, 0.018379757180809975, 0.05070207267999649, 0.02929568849503994, -0.019208213314414024, -0.052836835384368896, -0.057730354368686676, -0.008787300437688828, -0.0003151149139739573, -0.02382075972855091, ...
Proving Theorems using Incremental Learning and Hindsight Experience Replay
https://proceedings.mlr.press/v162/aygun22a.html
[ "Eser Aygün", "Ankit Anand", "Laurent Orseau", "Xavier Glorot", "Stephen M Mcaleer", "Vlad Firoiu", "Lei M Zhang", "Doina Precup", "Shibl Mourad" ]
null
null
Traditional automated theorem proving systems for first-order logic depend on speed-optimized search and many handcrafted heuristics designed to work over a wide range of domains. Machine learning approaches in the literature either depend on these traditional provers to bootstrap themselves, by leveraging these heuris...
[]
null
49
2112.10664
title_snapshot
[ -0.05149330198764801, -0.013501245528459549, -0.0025496785528957844, 0.04619424790143967, 0.04723450914025307, 0.01091453805565834, 0.029350224882364273, -0.02695806510746479, -0.024573082104325294, -0.011029321700334549, 0.015622353181242943, 0.05446535721421242, -0.0400981642305851, -0.0...
Near-optimal rate of consistency for linear models with missing values
https://proceedings.mlr.press/v162/ayme22a.html
[ "Alexis Ayme", "Claire Boyer", "Aymeric Dieuleveut", "Erwan Scornet" ]
null
null
Missing values arise in most real-world data sets due to the aggregation of multiple sources and intrinsically missing information (sensor failure, unanswered questions in surveys...). In fact, the very nature of missing values usually prevents us from running standard learning algorithms. In this paper, we focus on th...
[]
null
50
2202.01463
title_judge
[ -0.0011724171927198768, -0.014022411778569221, -0.02243395708501339, 0.05853146314620972, 0.054332759231328964, 0.057808857411146164, 0.0340086929500103, -0.007739597000181675, -0.04100586846470833, -0.04633459076285362, -0.035072337836027145, 0.014192112721502781, -0.07574380189180374, -0...
How Tempering Fixes Data Augmentation in Bayesian Neural Networks
https://proceedings.mlr.press/v162/bachmann22a.html
[ "Gregor Bachmann", "Lorenzo Noci", "Thomas Hofmann" ]
null
null
While Bayesian neural networks (BNNs) provide a sound and principled alternative to standard neural networks, an artificial sharpening of the posterior usually needs to be applied to reach comparable performance. This is in stark contrast to theory, dictating that given an adequate prior and a well-specified model, the...
[]
null
51
2205.13900
title_snapshot
[ -0.005185430403798819, 0.0033461018465459347, -0.029098540544509888, 0.0659954845905304, 0.056100450456142426, 0.03171800076961517, 0.033050134778022766, 0.009737671352922916, -0.037068068981170654, -0.06354652345180511, -0.01033997442573309, 0.03576062247157097, -0.05457635968923569, -0.0...
ASAP.SGD: Instance-based Adaptiveness to Staleness in Asynchronous SGD
https://proceedings.mlr.press/v162/backstrom22a.html
[ "Karl Bäckström", "Marina Papatriantafilou", "Philippas Tsigas" ]
null
null
Concurrent algorithmic implementations of Stochastic Gradient Descent (SGD) give rise to critical questions for compute-intensive Machine Learning (ML). Asynchrony implies speedup in some contexts, and challenges in others, as stale updates may lead to slower, or non-converging executions. While previous works showed a...
[]
null
52
null
null
[ -0.043230410665273666, -0.04320601746439934, -0.005332958418875933, 0.05228114500641823, 0.028008149936795235, 0.029284803196787834, 0.042872995138168335, 0.0069174082018435, -0.04370526596903801, -0.033471424132585526, -0.00463637663051486, -0.01969739980995655, -0.056886471807956696, -0....
From Noisy Prediction to True Label: Noisy Prediction Calibration via Generative Model
https://proceedings.mlr.press/v162/bae22a.html
[ "Heesun Bae", "Seungjae Shin", "Byeonghu Na", "Joonho Jang", "Kyungwoo Song", "Il-Chul Moon" ]
null
null
Noisy labels are inevitable yet problematic in machine learning society. It ruins the generalization of a classifier by making the classifier over-fitted to noisy labels. Existing methods on noisy label have focused on modifying the classifier during the training procedure. It has two potential problems. First, these m...
[]
null
53
2205.00690
title_snapshot
[ -0.00020963145652785897, -0.04526878520846367, -0.0343189500272274, 0.028455646708607674, 0.03419330343604088, 0.025829359889030457, 0.023289348930120468, -0.006315659265965223, -0.03710190951824188, -0.04456419497728348, -0.020369460806250572, 0.02757231518626213, -0.06521492451429367, 0....
data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language
https://proceedings.mlr.press/v162/baevski22a.html
[ "Alexei Baevski", "Wei-Ning Hsu", "Qiantong Xu", "Arun Babu", "Jiatao Gu", "Michael Auli" ]
null
null
While the general idea of self-supervised learning is identical across modalities, the actual algorithms and objectives differ widely because they were developed with a single modality in mind. To get us closer to general self-supervised learning, we present data2vec, a framework that uses the same learning method for ...
[]
null
54
2202.03555
title_snapshot
[ 0.003131850389763713, -0.03179387375712395, -0.013502324931323528, 0.05202211067080498, 0.011298081837594509, 0.06610194593667984, 0.04647662863135338, 0.0008147210464812815, -0.0050396788865327835, -0.024551182985305786, -0.0045898365788161755, 0.01998511329293251, -0.06801674515008926, 0...
End-to-End Balancing for Causal Continuous Treatment-Effect Estimation
https://proceedings.mlr.press/v162/bahadori22a.html
[ "Taha Bahadori", "Eric Tchetgen Tchetgen", "David Heckerman" ]
null
null
We study the problem of observational causal inference with continuous treatment. We focus on the challenge of estimating the causal response curve for infrequently-observed treatment values. We design a new algorithm based on the framework of entropy balancing which learns weights that directly maximize causal inferen...
[]
null
55
2107.13068
title_snapshot
[ 0.008944513276219368, -0.028585974127054214, -0.03546175733208656, 0.012755291536450386, 0.02878251112997532, 0.02487260103225708, 0.039111681282520294, -0.01115448772907257, -0.013208908028900623, -0.052960313856601715, -0.010687450878322124, 0.009711765684187412, -0.059994474053382874, -...
A Hierarchical Transitive-Aligned Graph Kernel for Un-attributed Graphs
https://proceedings.mlr.press/v162/bai22a.html
[ "Lu Bai", "Lixin Cui", "Hancock Edwin" ]
null
null
In this paper, we develop a new graph kernel, namely the Hierarchical Transitive-Aligned Kernel, by transitively aligning the vertices between graphs through a family of hierarchical prototype graphs. Comparing to most existing state-of-the-art graph kernels, the proposed kernel has three theoretical advantages. First,...
[]
null
56
2002.04425
title_snapshot
[ 0.002494588727131486, -0.0193721204996109, 0.04001708701252937, 0.053276125341653824, -0.004927005618810654, 0.03453534096479416, 0.0085749551653862, 0.02268395386636257, 0.0031939102336764336, -0.048393748700618744, -0.004199419170618057, -0.012483580969274044, -0.06888852268457413, 0.013...
Near-Optimal Learning of Extensive-Form Games with Imperfect Information
https://proceedings.mlr.press/v162/bai22b.html
[ "Yu Bai", "Chi Jin", "Song Mei", "Tiancheng Yu" ]
null
null
This paper resolves the open question of designing near-optimal algorithms for learning imperfect-information extensive-form games from bandit feedback. We present the first line of algorithms that require only $\widetilde{\mathcal{O}}((XA+YB)/\varepsilon^2)$ episodes of play to find an $\varepsilon$-approximate Nash e...
[]
null
57
2202.01752
title_snapshot
[ -0.041467051953077316, -0.02338387444615364, 0.0055436366237699986, 0.03744503855705261, 0.040081217885017395, 0.011429721489548683, 0.0003967406810261309, 0.013035845011472702, -0.017367742955684662, -0.045911625027656555, -0.013807017356157303, 0.01919718086719513, -0.053734730929136276, ...
Gaussian Mixture Variational Autoencoder with Contrastive Learning for Multi-Label Classification
https://proceedings.mlr.press/v162/bai22c.html
[ "Junwen Bai", "Shufeng Kong", "Carla P Gomes" ]
null
null
Multi-label classification (MLC) is a prediction task where each sample can have more than one label. We propose a novel contrastive learning boosted multi-label prediction model based on a Gaussian mixture variational autoencoder (C-GMVAE), which learns a multimodal prior space and employs a contrastive loss. Many exi...
[]
null
58
2112.00976
title_snapshot
[ 0.01737980730831623, -0.015822211280465126, 0.001640856615267694, 0.054411228746175766, 0.007625626865774393, 0.035749197006225586, 0.018141306936740875, 0.0010897855972871184, -0.04479743540287018, -0.029201561585068703, -0.007974524050951004, 0.015804272145032883, -0.07752399891614914, 0...
A$^3$T: Alignment-Aware Acoustic and Text Pretraining for Speech Synthesis and Editing
https://proceedings.mlr.press/v162/bai22d.html
[ "He Bai", "Renjie Zheng", "Junkun Chen", "Mingbo Ma", "Xintong Li", "Liang Huang" ]
null
null
Recently, speech representation learning has improved many speech-related tasks such as speech recognition, speech classification, and speech-to-text translation. However, all the above tasks are in the direction of speech understanding, but for the inverse direction, speech synthesis, the potential of representation l...
[]
null
59
2203.09690
title_snapshot
[ -0.005579183343797922, -0.002497578039765358, -0.03639134019613266, 0.03907241299748421, 0.016964778304100037, 0.05376499518752098, 0.04792637750506401, 0.038793522864580154, -0.036274414509534836, -0.038998860865831375, -0.019167957827448845, 0.024516170844435692, -0.05861569568514824, -0...
Stability Based Generalization Bounds for Exponential Family Langevin Dynamics
https://proceedings.mlr.press/v162/banerjee22a.html
[ "Arindam Banerjee", "Tiancong Chen", "Xinyan Li", "Yingxue Zhou" ]
null
null
Recent years have seen advances in generalization bounds for noisy stochastic algorithms, especially stochastic gradient Langevin dynamics (SGLD) based on stability (Mou et al., 2018; Li et al., 2020) and information theoretic approaches (Xu & Raginsky, 2017; Negrea et al., 2019; Steinke & Zakynthinou, 2020). In this p...
[]
null
60
2201.03064
title_snapshot
[ -0.006189387757331133, 0.004980870522558689, 0.012925814837217331, 0.021772829815745354, 0.05655023455619812, 0.024699199944734573, 0.05293549969792366, -0.0195610411465168, -0.03477765619754791, -0.04975714907050133, 0.01668635755777359, 0.0010749604552984238, -0.0944303423166275, -0.0233...
Certified Neural Network Watermarks with Randomized Smoothing
https://proceedings.mlr.press/v162/bansal22a.html
[ "Arpit Bansal", "Ping-Yeh Chiang", "Michael J Curry", "Rajiv Jain", "Curtis Wigington", "Varun Manjunatha", "John P Dickerson", "Tom Goldstein" ]
null
null
Watermarking is a commonly used strategy to protect creators’ rights to digital images, videos and audio. Recently, watermarking methods have been extended to deep learning models – in principle, the watermark should be preserved when an adversary tries to copy the model. However, in practice, watermarks can often be r...
[]
null
61
2207.07972
title_snapshot
[ 0.019651364535093307, -0.032283712178468704, -0.016902275383472443, 0.05437345430254936, 0.04748032987117767, 0.012682006694376469, 0.022561650723218918, -0.024653175845742226, -0.03809325397014618, -0.046629659831523895, -0.007005518767982721, -0.010901008732616901, -0.026155000552535057, ...
Data Scaling Laws in NMT: The Effect of Noise and Architecture
https://proceedings.mlr.press/v162/bansal22b.html
[ "Yamini Bansal", "Behrooz Ghorbani", "Ankush Garg", "Biao Zhang", "Colin Cherry", "Behnam Neyshabur", "Orhan Firat" ]
null
null
In this work, we study the effect of varying the architecture and training data quality on the data scaling properties of Neural Machine Translation (NMT). First, we establish that the test loss of encoder-decoder transformer models scales as a power law in the number of training samples, with a dependence on the model...
[]
null
62
2202.01994
title_snapshot
[ -0.04100368171930313, -0.02532743103802204, -0.01296835858374834, 0.028900587931275368, 0.031202757731080055, 0.05290955677628517, 0.02772059291601181, 0.010464297607541084, -0.0319700762629509, -0.006542562507092953, 0.00940136518329382, 0.02140011452138424, -0.060129426419734955, 0.01699...
Learning Stable Classifiers by Transferring Unstable Features
https://proceedings.mlr.press/v162/bao22a.html
[ "Yujia Bao", "Shiyu Chang", "Dr.Regina Barzilay" ]
null
null
While unbiased machine learning models are essential for many applications, bias is a human-defined concept that can vary across tasks. Given only input-label pairs, algorithms may lack sufficient information to distinguish stable (causal) features from unstable (spurious) features. However, related tasks often share s...
[]
null
63
2106.07847
title_snapshot
[ 0.006094538606703281, -0.014496040530502796, -0.018795737996697426, 0.028378959745168686, 0.03266369178891182, 0.011997007764875889, 0.004716828931123018, -0.011800533160567284, -0.01760437712073326, -0.05397946015000343, -0.014403006993234158, 0.008075905032455921, -0.09605500102043152, 0...
Fast Composite Optimization and Statistical Recovery in Federated Learning
https://proceedings.mlr.press/v162/bao22b.html
[ "Yajie Bao", "Michael Crawshaw", "Shan Luo", "Mingrui Liu" ]
null
null
As a prevalent distributed learning paradigm, Federated Learning (FL) trains a global model on a massive amount of devices with infrequent communication. This paper investigates a class of composite optimization and statistical recovery problems in the FL setting, whose loss function consists of a data-dependent smooth...
[]
null
64
2207.08204
title_snapshot
[ -0.022757474333047867, -0.04871363192796707, 0.02412399649620056, 0.041640639305114746, 0.033172521740198135, 0.01902918703854084, 0.01492868922650814, -0.014772318303585052, -0.025012770667672157, -0.027885133400559425, -0.03413126617670059, -0.01929803565144539, -0.029731471091508865, 0....
Generative Modeling for Multi-task Visual Learning
https://proceedings.mlr.press/v162/bao22c.html
[ "Zhipeng Bao", "Martial Hebert", "Yu-Xiong Wang" ]
null
null
Generative modeling has recently shown great promise in computer vision, but it has mostly focused on synthesizing visually realistic images. In this paper, motivated by multi-task learning of shareable feature representations, we consider a novel problem of learning a shared generative model that is useful across vari...
[]
null
65
2106.13409
title_snapshot
[ 0.0337679460644722, -0.02048513852059841, 0.0036759632639586926, 0.04632328450679779, 0.021410521119832993, 0.03118516318500042, 0.012729235924780369, 0.0287680272012949, -0.010770893655717373, -0.06100169196724892, -0.03332804888486862, 0.011238868348300457, -0.0806593969464302, -0.000890...
Estimating the Optimal Covariance with Imperfect Mean in Diffusion Probabilistic Models
https://proceedings.mlr.press/v162/bao22d.html
[ "Fan Bao", "Chongxuan Li", "Jiacheng Sun", "Jun Zhu", "Bo Zhang" ]
null
null
Diffusion probabilistic models (DPMs) are a class of powerful deep generative models (DGMs). Despite their success, the iterative generation process over the full timesteps is much less efficient than other DGMs such as GANs. Thus, the generation performance on a subset of timesteps is crucial, which is greatly influen...
[]
null
66
2206.07309
title_snapshot
[ -0.03981029614806175, -0.003912320826202631, -0.015082122758030891, 0.04150240868330002, 0.04208417981863022, 0.04861633852124214, 0.046847015619277954, -0.029551615938544273, 0.006171250715851784, -0.04714447259902954, -0.0027296198531985283, -0.03762541338801384, -0.05352133885025978, 0....
On the Surrogate Gap between Contrastive and Supervised Losses
https://proceedings.mlr.press/v162/bao22e.html
[ "Han Bao", "Yoshihiro Nagano", "Kento Nozawa" ]
null
null
Contrastive representation learning encourages data representation to make semantically similar pairs closer than randomly drawn negative samples, which has been successful in various domains such as vision, language, and graphs. Recent theoretical studies have attempted to explain the benefit of the large negative sam...
[]
null
67
2110.02501
title_snapshot
[ -0.024698512628674507, -0.004875452257692814, -0.015322946943342686, 0.05964512377977371, 0.02412990853190422, 0.015956444665789604, 0.044301506131887436, -0.000976976240053773, -0.04543415084481239, -0.016828037798404694, -0.012173040769994259, 0.009267505258321762, -0.07057841867208481, ...
Representation Topology Divergence: A Method for Comparing Neural Network Representations.
https://proceedings.mlr.press/v162/barannikov22a.html
[ "Serguei Barannikov", "Ilya Trofimov", "Nikita Balabin", "Evgeny Burnaev" ]
null
null
Comparison of data representations is a complex multi-aspect problem. We propose a method for comparing two data representations. We introduce the Representation Topology Divergence (RTD) score measuring the dissimilarity in multi-scale topology between two point clouds of equal size with a one-to-one correspondence be...
[]
null
68
2201.00058
title_snapshot
[ -0.03015241213142872, -0.028754225000739098, -0.009089975617825985, 0.03985302150249481, 0.034028809517621994, 0.02207602560520172, 0.018699193373322487, 0.004428991116583347, -0.04024876281619072, -0.034782495349645615, -0.009057795628905296, -0.006721656769514084, -0.06939004361629486, 0...
Sparse Mixed Linear Regression with Guarantees: Taming an Intractable Problem with Invex Relaxation
https://proceedings.mlr.press/v162/barik22a.html
[ "Adarsh Barik", "Jean Honorio" ]
null
null
In this paper, we study the problem of sparse mixed linear regression on an unlabeled dataset that is generated from linear measurements from two different regression parameter vectors. Since the data is unlabeled, our task is to not only figure out a good approximation of regression parameter vectors but also label th...
[]
null
69
2206.01167
title_snapshot
[ 0.0022678179666399956, -0.013568934984505177, -0.014919092878699303, 0.032116781920194626, 0.03008228912949562, 0.05561849847435951, 0.028760934248566628, -0.018824145197868347, -0.04086907207965851, -0.02469564788043499, -0.010151908732950687, -0.0059548052959144115, -0.064356230199337, -...
Neural Fisher Discriminant Analysis: Optimal Neural Network Embeddings in Polynomial Time
https://proceedings.mlr.press/v162/bartan22a.html
[ "Burak Bartan", "Mert Pilanci" ]
null
null
Fisher’s Linear Discriminant Analysis (FLDA) is a statistical analysis method that linearly embeds data points to a lower dimensional space to maximize a discrimination criterion such that the variance between classes is maximized while the variance within classes is minimized. We introduce a natural extension of FLDA ...
[]
null
70
null
null
[ -0.009793668054044247, -0.0033047397155314684, 0.0042008585296571255, 0.02176627889275551, 0.06311915814876556, 0.07512935250997543, -0.00810085516422987, -0.01399250142276287, -0.030000459402799606, -0.04782901704311371, 0.00785781815648079, 0.010458811186254025, -0.06231420114636421, -0....
Fictitious Play and Best-Response Dynamics in Identical Interest and Zero-Sum Stochastic Games
https://proceedings.mlr.press/v162/baudin22a.html
[ "Lucas Baudin", "Rida Laraki" ]
null
null
This paper proposes an extension of a popular decentralized discrete-time learning procedure when repeating a static game called fictitious play (FP) (Brown, 1951; Robinson, 1951) to a dynamic model called discounted stochastic game (Shapley, 1953). Our family of discrete-time FP procedures is proven to converge to the...
[]
null
71
2111.04317
title_judge
[ -0.04466133937239647, -0.01761416532099247, 0.012051926925778389, 0.021177390590310097, 0.031368959695100784, 0.033600322902202606, -0.00408285204321146, 0.045212700963020325, -0.03483116254210472, -0.043878596276044846, 0.02050531841814518, 0.004457233939319849, -0.07833541929721832, -0.0...
Information Discrepancy in Strategic Learning
https://proceedings.mlr.press/v162/bechavod22a.html
[ "Yahav Bechavod", "Chara Podimata", "Steven Wu", "Juba Ziani" ]
null
null
We initiate the study of the effects of non-transparency in decision rules on individuals’ ability to improve in strategic learning settings. Inspired by real-life settings, such as loan approvals and college admissions, we remove the assumption typically made in the strategic learning literature, that the decision rul...
[]
null
72
2103.01028
title_snapshot
[ -0.0352797769010067, -0.014729826711118221, -0.006577824242413044, 0.03754951059818268, 0.04863392934203148, -0.008738953620195389, 0.029707305133342743, 0.022355714812874794, -0.04895651713013649, -0.04233035817742348, -0.02447308413684368, 0.03330390900373459, -0.05476203188300133, -0.02...
On the Hidden Biases of Policy Mirror Ascent in Continuous Action Spaces
https://proceedings.mlr.press/v162/bedi22a.html
[ "Amrit Singh Bedi", "Souradip Chakraborty", "Anjaly Parayil", "Brian M Sadler", "Pratap Tokekar", "Alec Koppel" ]
null
null
We focus on parameterized policy search for reinforcement learning over continuous action spaces. Typically, one assumes the score function associated with a policy is bounded, which {fails to hold even for Gaussian policies. } To properly address this issue, one must introduce an exploration tolerance parameter to qua...
[]
null
73
2201.12332
title_snapshot
[ -0.026390867307782173, -0.01948929764330387, -0.008027937263250351, 0.0477278046309948, 0.023325549438595772, 0.005933302454650402, 0.038857217878103256, -0.025622202083468437, -0.015875669196248055, -0.027480298653244972, -0.01122782751917839, 0.005261797457933426, -0.07175307720899582, -...
Imitation Learning by Estimating Expertise of Demonstrators
https://proceedings.mlr.press/v162/beliaev22a.html
[ "Mark Beliaev", "Andy Shih", "Stefano Ermon", "Dorsa Sadigh", "Ramtin Pedarsani" ]
null
null
Many existing imitation learning datasets are collected from multiple demonstrators, each with different expertise at different parts of the environment. Yet, standard imitation learning algorithms typically treat all demonstrators as homogeneous, regardless of their expertise, absorbing the weaknesses of any suboptima...
[]
null
74
2202.01288
title_snapshot
[ -0.010973027907311916, -0.03501926735043526, -0.014148910529911518, 0.0442923866212368, 0.047287654131650925, 0.020658768713474274, 0.01919894479215145, 0.014528555795550346, -0.032868094742298126, -0.019618920981884003, 0.0010420226026326418, 0.003327291924506426, -0.07055418193340302, -0...
Matching Normalizing Flows and Probability Paths on Manifolds
https://proceedings.mlr.press/v162/ben-hamu22a.html
[ "Heli Ben-Hamu", "Samuel Cohen", "Joey Bose", "Brandon Amos", "Maximillian Nickel", "Aditya Grover", "Ricky T. Q. Chen", "Yaron Lipman" ]
null
null
Continuous Normalizing Flows (CNFs) are a class of generative models that transform a prior distribution to a model distribution by solving an ordinary differential equation (ODE). We propose to train CNFs on manifolds by minimizing probability path divergence (PPD), a novel family of divergences between the probabilit...
[]
null
75
2207.04711
title_snapshot
[ -0.008653301745653152, -0.0351831279695034, 0.008713401854038239, 0.061941247433423996, 0.04901282489299774, 0.04651210084557533, 0.00668323365971446, 0.002626829082146287, -0.014854904264211655, -0.06258704513311386, 0.00426270067691803, -0.00854007713496685, -0.05199143663048744, 0.00334...
Stochastic Contextual Dueling Bandits under Linear Stochastic Transitivity Models
https://proceedings.mlr.press/v162/bengs22a.html
[ "Viktor Bengs", "Aadirupa Saha", "Eyke Hüllermeier" ]
null
null
We consider the regret minimization task in a dueling bandits problem with context information. In every round of the sequential decision problem, the learner makes a context-dependent selection of two choice alternatives (arms) to be compared with each other and receives feedback in the form of noisy preference inform...
[]
null
76
2202.04593
title_snapshot
[ -0.027582867071032524, 0.01670660451054573, -0.007036114111542702, 0.051508404314517975, 0.005432222969830036, 0.024920953437685966, 0.02326316200196743, 0.03537534922361374, -0.015430551953613758, -0.045893263071775436, -0.02583444118499756, 0.027541209012269974, -0.05295025184750557, -0....
Neural Inverse Kinematic
https://proceedings.mlr.press/v162/bensadoun22a.html
[ "Raphael Bensadoun", "Shir Gur", "Nitsan Blau", "Lior Wolf" ]
null
null
Inverse kinematic (IK) methods recover the parameters of the joints, given the desired position of selected elements in the kinematic chain. While the problem is well-defined and low-dimensional, it has to be solved rapidly, accounting for multiple possible solutions. In this work, we propose a neural IK method that em...
[]
null
77
null
null
[ -0.03056289628148079, -0.0010817856527864933, -0.01693398505449295, -0.004295818507671356, 0.03727501630783081, 0.05704737827181816, 0.022245565429329872, -0.02427327260375023, -0.02393379434943199, -0.05624932795763016, 0.03355010971426964, -0.029200466349720955, -0.038239095360040665, -0...
Volatility Based Kernels and Moving Average Means for Accurate Forecasting with Gaussian Processes
https://proceedings.mlr.press/v162/benton22a.html
[ "Gregory Benton", "Wesley Maddox", "Andrew Gordon Wilson" ]
null
null
A broad class of stochastic volatility models are defined by systems of stochastic differential equations, and while these models have seen widespread success in domains such as finance and statistical climatology, they typically lack an ability to condition on historical data to produce a true posterior distribution. ...
[]
null
78
2207.06544
title_snapshot
[ -0.026933865621685982, -0.02831091172993183, 0.021728985011577606, 0.018200915306806564, 0.0313400961458683, 0.04454837366938591, 0.038343645632267, 0.008507334627211094, -0.03293183445930481, -0.050610508769750595, 0.005933200940489769, 0.019148193299770355, -0.0626542940735817, 0.0209828...
Gradient Descent on Neurons and its Link to Approximate Second-order Optimization
https://proceedings.mlr.press/v162/benzing22a.html
[ "Frederik Benzing" ]
null
null
Second-order optimizers are thought to hold the potential to speed up neural network training, but due to the enormous size of the curvature matrix, they typically require approximations to be computationally tractable. The most successful family of approximations are Kronecker-Factored, block-diagonal curvature estima...
[]
null
79
2201.12250
title_snapshot
[ -0.042149513959884644, -0.02173231914639473, 0.018093015998601913, 0.017123710364103317, 0.029794052243232727, 0.038721296936273575, 0.03217763453722, 0.007975786924362183, -0.035718753933906555, -0.056034114211797714, 0.0013738212874159217, -0.011910829693078995, -0.05837101861834526, 0.0...
Safe Learning in Tree-Form Sequential Decision Making: Handling Hard and Soft Constraints
https://proceedings.mlr.press/v162/bernasconi22a.html
[ "Martino Bernasconi", "Federico Cacciamani", "Matteo Castiglioni", "Alberto Marchesi", "Nicola Gatti", "Francesco Trovò" ]
null
null
We study decision making problems in which an agent sequentially interacts with a stochastic environment defined by means of a tree structure. The agent repeatedly faces the environment over time, and, after each round, it perceives a utility and a cost, which are both stochastic. The goal of the agent is to learn an o...
[]
null
80
null
null
[ -0.057302091270685196, -0.012674588710069656, 0.005552539601922035, 0.022437818348407745, 0.0356290228664875, 0.013166315853595734, 0.02676672674715519, 0.02539178356528282, -0.01639200933277607, -0.0355987511575222, -0.018857697024941444, 0.017222026363015175, -0.05753067135810852, -0.040...
Skin Deep Unlearning: Artefact and Instrument Debiasing in the Context of Melanoma Classification
https://proceedings.mlr.press/v162/bevan22a.html
[ "Peter Bevan", "Amir Atapour-Abarghouei" ]
null
null
Convolutional Neural Networks have demonstrated dermatologist-level performance in the classification of melanoma from skin lesion images, but prediction irregularities due to biases seen within the training data are an issue that should be addressed before widespread deployment is possible. In this work, we robustly r...
[]
null
81
2109.09818
title_snapshot
[ 0.0124961007386446, -0.041648030281066895, -0.011550425551831722, -0.010837160050868988, 0.04270067438483238, -0.016484592109918594, 0.03783430531620979, -0.0027955605182796717, -0.02251856029033661, -0.061374541372060776, 0.007150443736463785, 0.04246765747666359, -0.06539050489664078, 0....
Approximate Bayesian Computation with Domain Expert in the Loop
https://proceedings.mlr.press/v162/bharti22a.html
[ "Ayush Bharti", "Louis Filstroff", "Samuel Kaski" ]
null
null
Approximate Bayesian computation (ABC) is a popular likelihood-free inference method for models with intractable likelihood functions. As ABC methods usually rely on comparing summary statistics of observed and simulated data, the choice of the statistics is crucial. This choice involves a trade-off between loss of inf...
[]
null
82
2201.12090
title_snapshot
[ -0.005655521526932716, 0.011644164100289345, -0.01650911197066307, 0.004345445893704891, 0.044973891228437424, 0.014894744381308556, 0.04289001226425171, -0.05085994675755501, -0.02403428591787815, -0.04173871502280235, -0.017676951363682747, 0.014258768409490585, -0.07914617657661438, 0.0...
Minimax M-estimation under Adversarial Contamination
https://proceedings.mlr.press/v162/bhatt22a.html
[ "Sujay Bhatt", "Guanhua Fang", "Ping Li", "Gennady Samorodnitsky" ]
null
null
We present a new finite-sample analysis of Catoni’s M-estimator under adversarial contamination, where an adversary is allowed to corrupt a fraction of the samples arbitrarily. We make minimal assumptions on the distribution of the uncontaminated random variables, namely, we only assume the existence of a known upper b...
[]
null
83
null
null
[ -0.027547912672162056, -0.0020244098268449306, -0.007072058971971273, 0.055894553661346436, 0.04452573135495186, 0.03175073862075806, 0.036196865141391754, -0.010394317097961903, -0.030821142718195915, -0.04674569517374039, 0.006052384153008461, -0.00784173235297203, -0.07632356137037277, ...
Nearly Optimal Catoni’s M-estimator for Infinite Variance
https://proceedings.mlr.press/v162/bhatt22b.html
[ "Sujay Bhatt", "Guanhua Fang", "Ping Li", "Gennady Samorodnitsky" ]
null
null
In this paper, we extend the remarkable M-estimator of Catoni \citep{Cat12} to situations where the variance is infinite. In particular, given a sequence of i.i.d random variables $\{X_i\}_{i=1}^n$ from distribution $\mathcal{D}$ over $\mathbb{R}$ with mean $\mu$, we only assume the existence of a known upper bound $\u...
[]
null
84
null
null
[ -0.03356931358575821, -0.019270408898591995, -0.004515402484685183, 0.017161335796117783, 0.05909835919737816, 0.0318765714764595, 0.03852133825421333, 0.02238035574555397, -0.0318806916475296, -0.052126236259937286, 0.021944595500826836, -0.022036166861653328, -0.0632108747959137, 0.01008...
Personalization Improves Privacy-Accuracy Tradeoffs in Federated Learning
https://proceedings.mlr.press/v162/bietti22a.html
[ "Alberto Bietti", "Chen-Yu Wei", "Miroslav Dudik", "John Langford", "Steven Wu" ]
null
null
Large-scale machine learning systems often involve data distributed across a collection of users. Federated learning algorithms leverage this structure by communicating model updates to a central server, rather than entire datasets. In this paper, we study stochastic optimization algorithms for a personalized federated...
[]
null
85
2202.05318
title_snapshot
[ 0.01069728098809719, -0.00013795866107102484, 0.008683125488460064, 0.07679639011621475, 0.06590144336223602, 0.0165510643273592, 0.03750437870621681, -0.020716164261102676, -0.002793524181470275, -0.027755126357078552, -0.002045117085799575, -0.008580002933740616, -0.033841244876384735, 0...
Non-Vacuous Generalisation Bounds for Shallow Neural Networks
https://proceedings.mlr.press/v162/biggs22a.html
[ "Felix Biggs", "Benjamin Guedj" ]
null
null
We focus on a specific class of shallow neural networks with a single hidden layer, namely those with $L_2$-normalised data and either a sigmoid-shaped Gaussian error function (“erf”) activation or a Gaussian Error Linear Unit (GELU) activation. For these networks, we derive new generalisation bounds through the PAC-Ba...
[]
null
86
2202.01627
title_snapshot
[ -0.02447780780494213, -0.01592472940683365, -0.012409971095621586, 0.04700884595513344, 0.01983131282031536, 0.0670827329158783, 0.04828915372490883, -0.005660422146320343, -0.05113185942173004, -0.028868891298770905, -0.005424363538622856, 0.007756248582154512, -0.06384661793708801, 0.000...
Structure-preserving GANs
https://proceedings.mlr.press/v162/birrell22a.html
[ "Jeremiah Birrell", "Markos Katsoulakis", "Luc Rey-Bellet", "Wei Zhu" ]
null
null
Generative adversarial networks (GANs), a class of distribution-learning methods based on a two-player game between a generator and a discriminator, can generally be formulated as a minmax problem based on the variational representation of a divergence between the unknown and the generated distributions. We introduce s...
[]
null
87
2202.01129
title_snapshot
[ -0.003984859213232994, -0.007826156914234161, 0.005936740897595882, 0.041363656520843506, 0.014943009242415428, 0.00849439762532711, 0.0033049762714654207, -0.016241569072008133, -0.0007483638473786414, -0.06511993706226349, 0.017558321356773376, -0.009592011570930481, -0.0658128410577774, ...
Scalable Spike-and-Slab
https://proceedings.mlr.press/v162/biswas22a.html
[ "Niloy Biswas", "Lester Mackey", "Xiao-Li Meng" ]
null
null
Spike-and-slab priors are commonly used for Bayesian variable selection, due to their interpretability and favorable statistical properties. However, existing samplers for spike-and-slab posteriors incur prohibitive computational costs when the number of variables is large. In this article, we propose Scalable Spike-an...
[]
null
88
2204.01668
title_snapshot
[ -0.013693714514374733, -0.016386866569519043, -0.021680422127246857, 0.032454803586006165, 0.03018324449658394, 0.02536896988749504, 0.02426716685295105, -0.016932496801018715, -0.01281981635838747, -0.07584603130817413, 0.01281316764652729, -0.04037290811538696, -0.05085291713476181, 0.00...
Breaking Down Out-of-Distribution Detection: Many Methods Based on OOD Training Data Estimate a Combination of the Same Core Quantities
https://proceedings.mlr.press/v162/bitterwolf22a.html
[ "Julian Bitterwolf", "Alexander Meinke", "Maximilian Augustin", "Matthias Hein" ]
null
null
It is an important problem in trustworthy machine learning to recognize out-of-distribution (OOD) inputs which are inputs unrelated to the in-distribution task. Many out-of-distribution detection methods have been suggested in recent years. The goal of this paper is to recognize common objectives as well as to identify...
[]
null
89
2206.09880
title_snapshot
[ 0.0063428571447730064, -0.005228028632700443, -0.004638112150132656, 0.03892531618475914, 0.05422699451446533, 0.005552398040890694, 0.0165514275431633, -0.0002746226091403514, -0.02742345817387104, -0.021035611629486084, -0.004190817940980196, 0.01455832738429308, -0.08641701936721802, -0...
A query-optimal algorithm for finding counterfactuals
https://proceedings.mlr.press/v162/blanc22a.html
[ "Guy Blanc", "Caleb Koch", "Jane Lange", "Li-Yang Tan" ]
null
null
We design an algorithm for finding counterfactuals with strong theoretical guarantees on its performance. For any monotone model $f : X^d \to \{0,1\}$ and instance $x^\star$, our algorithm makes \[{S}(f)^{O(\Delta_f(x^\star))}\cdot \log d\]{queries} to $f$ and returns an {\sl optimal} counterfactual for $x^\star$: a ne...
[]
null
90
2207.07072
title_snapshot
[ -0.05275224521756172, 0.005110261030495167, 0.018391378223896027, 0.027488412335515022, 0.0441155843436718, 0.016942230984568596, 0.03719319775700569, 0.0060255154967308044, -0.022357938811182976, -0.02755209431052208, -0.0262916162610054, 0.03801758587360382, -0.07498542219400406, -0.0110...
Popular decision tree algorithms are provably noise tolerant
https://proceedings.mlr.press/v162/blanc22b.html
[ "Guy Blanc", "Jane Lange", "Ali Malik", "Li-Yang Tan" ]
null
null
Using the framework of boosting, we prove that all impurity-based decision tree learning algorithms, including the classic ID3, C4.5, and CART, are highly noise tolerant. Our guarantees hold under the strongest noise model of nasty noise, and we provide near-matching upper and lower bounds on the allowable noise rate. ...
[]
null
91
2206.08899
title_snapshot
[ -0.01225149817764759, -0.021841317415237427, -0.026788370683789253, 0.06080813705921173, 0.05109166353940964, 0.01304380502551794, 0.04337809607386589, -0.007449926808476448, 0.0033065287861973047, -0.019740529358386993, -0.013633901253342628, 0.008932693861424923, -0.06244538351893425, -0...
Optimizing Sequential Experimental Design with Deep Reinforcement Learning
https://proceedings.mlr.press/v162/blau22a.html
[ "Tom Blau", "Edwin V. Bonilla", "Iadine Chades", "Amir Dezfouli" ]
null
null
Bayesian approaches developed to solve the optimal design of sequential experiments are mathematically elegant but computationally challenging. Recently, techniques using amortization have been proposed to make these Bayesian approaches practical, by training a parameterized policy that proposes designs efficiently at ...
[]
null
92
2202.00821
title_snapshot
[ -0.01074319239705801, 0.0005189896910451353, -0.03435332700610161, 0.07406678795814514, 0.06607437878847122, 0.02577604539692402, 0.019222818315029144, -0.019015982747077942, 0.021173760294914246, -0.05793134495615959, -0.010534025728702545, 0.01959388703107834, -0.055198077112436295, -0.0...
Lagrangian Method for Q-Function Learning (with Applications to Machine Translation)
https://proceedings.mlr.press/v162/bojun22a.html
[ "Huang Bojun" ]
null
null
This paper discusses a new approach to the fundamental problem of learning optimal Q-functions. In this approach, optimal Q-functions are formulated as saddle points of a nonlinear Lagrangian function derived from the classic Bellman optimality equation. The paper shows that the Lagrangian enjoys strong duality, in spi...
[]
null
93
2207.11161
title_snapshot
[ -0.030039401724934578, -0.013141527771949768, -0.0041204774752259254, 0.00981643982231617, 0.058645810931921005, 0.04810159280896187, 0.019682012498378754, -0.0020368825644254684, -0.03551390767097473, -0.027092894539237022, -0.017777511849999428, 0.03291945159435272, -0.06682389229536057, ...
Generalized Results for the Existence and Consistency of the MLE in the Bradley-Terry-Luce Model
https://proceedings.mlr.press/v162/bong22a.html
[ "Heejong Bong", "Alessandro Rinaldo" ]
null
null
Ranking problems based on pairwise comparisons, such as those arising in online gaming, often involve a large pool of items to order. In these situations, the gap in performance between any two items can be significant, and the smallest and largest winning probabilities can be very close to zero or one. Furthermore, ea...
[]
null
94
2110.11487
title_snapshot
[ -0.014062467962503433, -0.011672557331621647, 0.00567884324118495, 0.03400657698512077, 0.04020491614937782, 0.0008373678429052234, 0.011993312276899815, 0.042803388088941574, -0.04366414248943329, -0.04353810101747513, -0.013662798330187798, 0.01458465401083231, -0.0735202506184578, -0.00...
How to Train Your Wide Neural Network Without Backprop: An Input-Weight Alignment Perspective
https://proceedings.mlr.press/v162/boopathy22a.html
[ "Akhilan Boopathy", "Ila Fiete" ]
null
null
Recent works have examined theoretical and empirical properties of wide neural networks trained in the Neural Tangent Kernel (NTK) regime. Given that biological neural networks are much wider than their artificial counterparts, we consider NTK regime wide neural networks as a possible model of biological neural network...
[]
null
95
2106.08453
title_snapshot
[ -0.03799687325954437, -0.034150756895542145, -0.019170012325048447, 0.021915854886174202, 0.037452805787324905, 0.013556593097746372, 0.017225895076990128, 0.020444408059120178, -0.020297685638070107, -0.027726728469133377, 0.016018731519579887, 0.031026219949126244, -0.07386250793933868, ...
Improving Language Models by Retrieving from Trillions of Tokens
https://proceedings.mlr.press/v162/borgeaud22a.html
[ "Sebastian Borgeaud", "Arthur Mensch", "Jordan Hoffmann", "Trevor Cai", "Eliza Rutherford", "Katie Millican", "George Bm Van Den Driessche", "Jean-Baptiste Lespiau", "Bogdan Damoc", "Aidan Clark", "Diego De Las Casas", "Aurelia Guy", "Jacob Menick", "Roman Ring", "Tom Hennigan", "Saffr...
null
null
We enhance auto-regressive language models by conditioning on document chunks retrieved from a large corpus, based on local similarity with preceding tokens. With a 2 trillion token database, our Retrieval-Enhanced Transformer (RETRO) obtains comparable performance to GPT-3 and Jurassic-1 on the Pile, despite using 25{...
[]
null
96
2112.04426
title_snapshot
[ -0.009034874849021435, -0.030589310452342033, -0.015785809606313705, 0.035722766071558, 0.019306717440485954, 0.03333970531821251, 0.004704310093075037, 0.018822532147169113, -0.015613982453942299, 0.0033211747650057077, -0.03914618119597435, 0.004281225148588419, -0.05061234161257744, -0....
Lie Point Symmetry Data Augmentation for Neural PDE Solvers
https://proceedings.mlr.press/v162/brandstetter22a.html
[ "Johannes Brandstetter", "Max Welling", "Daniel E Worrall" ]
null
null
Neural networks are increasingly being used to solve partial differential equations (PDEs), replacing slower numerical solvers. However, a critical issue is that neural PDE solvers require high-quality ground truth data, which usually must come from the very solvers they are designed to replace. Thus, we are presented ...
[]
null
97
2202.07643
title_snapshot
[ -0.060703348368406296, -0.005343617871403694, 0.018644139170646667, 0.04638637229800224, 0.02068309672176838, 0.03593454882502556, 0.015421232208609581, -0.03422948718070984, -0.032915595918893814, -0.06293688714504242, -0.0013711751671507955, -0.04031338915228844, -0.06570680439472198, 0....
An iterative clustering algorithm for the Contextual Stochastic Block Model with optimality guarantees
https://proceedings.mlr.press/v162/braun22a.html
[ "Guillaume Braun", "Hemant Tyagi", "Christophe Biernacki" ]
null
null
Real-world networks often come with side information that can help to improve the performance of network analysis tasks such as clustering. Despite a large number of empirical and theoretical studies conducted on network clustering methods during the past decade, the added value of side information and the methods used...
[]
null
98
2112.10467
title_snapshot
[ -0.010447047650814056, -0.02205042354762554, 0.0044351848773658276, 0.03448868915438652, 0.03711225092411041, 0.03885051608085632, 0.02728947252035141, 0.012095170095562935, -0.025696810334920883, -0.03475608304142952, 0.0123203219845891, -0.020998448133468628, -0.05252288281917572, -0.005...
Tractable Dendritic RNNs for Reconstructing Nonlinear Dynamical Systems
https://proceedings.mlr.press/v162/brenner22a.html
[ "Manuel Brenner", "Florian Hess", "Jonas M Mikhaeil", "Leonard F Bereska", "Zahra Monfared", "Po-Chen Kuo", "Daniel Durstewitz" ]
null
null
In many scientific disciplines, we are interested in inferring the nonlinear dynamical system underlying a set of observed time series, a challenging task in the face of chaotic behavior and noise. Previous deep learning approaches toward this goal often suffered from a lack of interpretability and tractability. In par...
[]
null
99
2207.02542
title_snapshot
[ -0.043914441019296646, -0.03095671907067299, -0.00768931582570076, 0.05305752903223038, 0.01316745299845934, 0.07059872150421143, 0.041239649057388306, 0.013294600881636143, -0.06018044427037239, -0.033857349306344986, 0.018102463334798813, -0.016901379451155663, -0.0440538115799427, -0.00...
Learning to Predict Graphs with Fused Gromov-Wasserstein Barycenters
https://proceedings.mlr.press/v162/brogat-motte22a.html
[ "Luc Brogat-Motte", "Rémi Flamary", "Celine Brouard", "Juho Rousu", "Florence D’Alché-Buc" ]
null
null
This paper introduces a novel and generic framework to solve the flagship task of supervised labeled graph prediction by leveraging Optimal Transport tools. We formulate the problem as regression with the Fused Gromov-Wasserstein (FGW) loss and propose a predictive model relying on a FGW barycenter whose weights depend...
[]
null
100
2202.03813
title_snapshot
[ -0.03637009859085083, -0.027999045327305794, 0.01610526815056801, 0.027207273989915848, 0.04181820899248123, 0.026114391162991524, 0.019689958542585373, 0.016708310693502426, 0.0018460986902937293, -0.06059548631310463, 0.019094690680503845, 0.0036426368169486523, -0.056757185608148575, 0....