COLT 2022

153 papers

Top Research Directions

Research directions most represented by COLT 2022 papers.

The Statistical Computational Gap in High Dimensions 15
Statistical Indistinguishability from Moment Matching 6
Multi-Agent Learning Under Bandit Feedback 5
The Boundaries of Statistical Query Hardness 4
High Probability Analysis for Reliable Stochastic Optimization 4
Regret Scaling in High Dimensional Online Learning 4
Dynamic Regret Minimization in Non-Stationary Online Learning 4
The Function Approximation Gap in Reinforcement Learning Theory 4
The Mechanics of Benign Overfitting in High Dimensions 4
Computational Lower Bounds via Statistical Queries 3
Sampling Beyond Log Concavity via Functional Inequalities 3
Extending Neural Network Theory Beyond Squared Loss 3
Proportional Limit Scaling in High Dimensional Learning 3
Agnostic Learning with Adversarial Label Noise 3
Dimension Efficient High Dimensional Private Estimation 3

Top Authors

AuthorPapers
Daniel Kane6
Ilias Diakonikolas5
Sinho Chewi4
Aaron Sidford3
Gautam Kamath3
Haipeng Luo3
Jason D. Lee3
Murat A Erdogdu3
Nicolas Flammarion3
Peter Bartlett3
Satyen Kale3
Tomer Koren3
Abhradeep Guha Thakurta2
Adil Salim2
Akshay Krishnamurthy2
Alekh Agarwal2
Alessandro Rudi2
Alexander Rakhlin2
Allen Liu2
Ankur Moitra2

Impactful Papers

Papers from COLT 2022 that shaped research clusters.

Foundational

Influential

Papers

Title
Private Robust Estimation by Stabilizing Convex Relaxations
Pushing the Efficiency-Regret Pareto Frontier for Online Learning of Portfolios and Quantum States
Random Graph Matching in Geometric Models: the Case of Complete Graphs
Rate of Convergence of Polynomial Networks to Gaussian Processes
Realizable Learning is All You Need
Return of the bias: Almost minimax optimal high probability bounds for adversarial linear bandits
Risk bounds for aggregated shallow neural networks using Gaussian priors
Robust Estimation for Random Graphs
Robustly-reliable learners under poisoning attacks
Robust Sparse Mean Estimation via Sum of Squares
ROOT-SGD: Sharp Nonasymptotics and Asymptotic Efficiency in a Single Algorithm
Sample-Efficient Reinforcement Learning in the Presence of Exogenous Information
Sampling Approximately Low-Rank Ising Models: MCMC meets Variational Methods
Scale-free Unconstrained Online Learning for Curved Losses
Self-Consistency of the Fokker Planck Equation
Sharp Constants in Uniformity Testing via the Huber Statistic
Sharper Rates for Separable Minimax and Finite Sum Optimization via Primal-Dual Extragradient Methods
Single Trajectory Nonparametric Learning of Nonlinear Dynamics
Smoothed Online Learning is as Easy as Statistical Learning
Stability vs Implicit Bias of Gradient Methods on Separable Data and Beyond
Statistical and Computational Phase Transitions in Group Testing
Statistical Estimation and Online Inference via Local SGD
Stochastic linear optimization never overfits with quadratically-bounded losses on general data
Stochastic Variance Reduction for Variational Inequality Methods
Strategizing against Learners in Bayesian Games
Streaming Algorithms for Ellipsoidal Approximation of Convex Polytopes
Strong Gaussian Approximation for the Sum of Random Vectors
Strong Memory Lower Bounds for Learning Natural Models
The Implicit Bias of Benign Overfitting
The merged-staircase property: a necessary and nearly sufficient condition for SGD learning of sparse functions on two-layer neural networks
The Pareto Frontier of Instance-Dependent Guarantees in Multi-Player Multi-Armed Bandits with no Communication
The Power of Adaptivity in SGD: Self-Tuning Step Sizes with Unbounded Gradients and Affine Variance
The Price of Tolerance in Distribution Testing
The Query Complexity of Local Search and Brouwer in Rounds
The query complexity of sampling from strongly log-concave distributions in one dimension
The Role of Interactivity in Structured Estimation
The Structured Abstain Problem and the Lovász Hinge
Thompson Sampling Achieves $\tilde{O}(\sqrt{T})$ Regret in Linear Quadratic Control
Tight query complexity bounds for learning graph partitions
Toward Instance-Optimal State Certification With Incoherent Measurements
Towards a Theory of Non-Log-Concave Sampling: First-Order Stationarity Guarantees for Langevin Monte Carlo
Towards Optimal Algorithms for Multi-Player Bandits without Collision Sensing Information
Trace norm regularization for multi-task learning with scarce data
Tracking Most Significant Arm Switches in Bandits
Two-Sided Weak Submodularity for Matroid Constrained Optimization and Regression
Understanding Riemannian Acceleration via a Proximal Extragradient Framework
Uniform Stability for First-Order Empirical Risk Minimization
Universality of empirical risk minimization
Universal Online Learning: an Optimistically Universal Learning Rule
Universal Online Learning with Bounded Loss: Reduction to Binary Classification

« prev Page 3/4 next »