English
Related papers

Related papers: Information-Theoretic Minimax Regret Bounds for Re…

200 papers

We study the regret of reinforcement learning from offline data generated by a fixed behavior policy in an infinite-horizon discounted Markov decision process (MDP). While existing analyses of common approaches, such as fitted $Q$-iteration…

Machine Learning · Computer Science 2023-07-13 Yichun Hu , Nathan Kallus , Masatoshi Uehara

Meta reinforcement learning sets a distribution over a set of tasks on which the agent can train at will, then is asked to learn an optimal policy for any test task efficiently. In this paper, we consider a finite set of tasks modeled…

Machine Learning · Computer Science 2024-06-05 Mirco Mutti , Aviv Tamar

We study how a decision-maker (DM) learns from data of unknown quality to form robust, ''general-purpose'' posterior beliefs. We develop a framework for robust learning and belief formation under a minimax-regret criterion, cast as a…

Theoretical Economics · Economics 2026-02-18 Yeon-Koo Che , Longjian Li , Tianling Luo

In this paper, we study the episodic reinforcement learning (RL) problem modeled by finite-horizon Markov Decision Processes (MDPs) with constraint on the number of batches. The multi-batch reinforcement learning framework, where the agent…

Machine Learning · Computer Science 2022-10-18 Zihan Zhang , Yuhang Jiang , Yuan Zhou , Xiangyang Ji

This is a brief technical note to clarify the state of lower bounds on regret for reinforcement learning. In particular, this paper: - Reproduces a lower bound on regret for reinforcement learning, similar to the result of Theorem 5 in the…

Machine Learning · Statistics 2016-08-10 Ian Osband , Benjamin Van Roy

We study regret minimization for infinite-horizon average-reward Markov Decision Processes (MDPs) under cost constraints. We start by designing a policy optimization algorithm with carefully designed action-value estimator and bonus term,…

Machine Learning · Computer Science 2022-02-02 Liyu Chen , Rahul Jain , Haipeng Luo

Modern tasks in reinforcement learning have large state and action spaces. To deal with them efficiently, one often uses predefined feature mapping to represent states and actions in a low-dimensional space. In this paper, we study…

Machine Learning · Computer Science 2021-02-24 Dongruo Zhou , Jiafan He , Quanquan Gu

We study finite episodic Markov decision processes incorporating dynamic risk measures to capture risk sensitivity. To this end, we present two model-based algorithms applied to \emph{Lipschitz} dynamic risk measures, a wide range of risk…

Machine Learning · Computer Science 2023-06-06 Hao Liang , Zhi-quan Luo

Information-theoretic Bayesian regret bounds of Russo and Van Roy capture the dependence of regret on prior uncertainty. However, this dependence is through entropy, which can become arbitrarily large as the number of actions increases. We…

Machine Learning · Statistics 2020-07-09 Shi Dong , Benjamin Van Roy

We consider reinforcement learning (RL) in episodic MDPs with adversarial full-information reward feedback and unknown fixed transition kernels. We propose two model-free policy optimization algorithms, POWER and POWER++, and establish…

Machine Learning · Computer Science 2020-07-02 Yingjie Fei , Zhuoran Yang , Zhaoran Wang , Qiaomin Xie

Policy optimization methods are popular reinforcement learning algorithms in practice. Recent works have built theoretical foundation for them by proving $\sqrt{T}$ regret bounds even when the losses are adversarial. Such bounds are tight…

Machine Learning · Computer Science 2023-02-21 Christoph Dann , Chen-Yu Wei , Julian Zimmert

A crucial problem in reinforcement learning is learning the optimal policy. We study this in tabular infinite-horizon discounted Markov decision processes under the online setting. The existing algorithms either fail to achieve regret…

Machine Learning · Computer Science 2023-12-13 Xiang Ji , Gen Li

We study the model-based undiscounted reinforcement learning for partially observable Markov decision processes (POMDPs). The oracle we consider is the optimal policy of the POMDP with a known environment in terms of the average reward over…

Machine Learning · Computer Science 2022-07-19 Yi Xiong , Ningyuan Chen , Xuefeng Gao , Xiang Zhou

This paper initiates the study of data-dependent regret bounds in constrained MAB settings. These bounds depend on the sequence of losses that characterize the problem instance. Thus, they can be much smaller than classical…

Machine Learning · Computer Science 2025-06-24 Gianmarco Genalti , Francesco Emanuele Stradi , Matteo Castiglioni , Alberto Marchesi , Nicola Gatti

Throughout the cognitive-science literature, there is widespread agreement that decision-making agents operating in the real world do so under limited information-processing capabilities and without access to unbounded cognitive or…

Machine Learning · Computer Science 2022-11-01 Dilip Arumugam , Mark K. Ho , Noah D. Goodman , Benjamin Van Roy

We study reinforcement learning with delayed state observation, where the agent observes the current state after some random number of time steps. We propose an algorithm that combines the augmentation method and the upper confidence bound…

Machine Learning · Computer Science 2026-03-05 Harin Lee , Kevin Jamieson

Constrained Markov decision processes (CMDPs) are a common way to model safety constraints in reinforcement learning. State-of-the-art methods for efficiently solving CMDPs are based on primal-dual algorithms. For these algorithms, all…

Machine Learning · Computer Science 2024-07-22 Adrian Müller , Pragnya Alatur , Volkan Cevher , Giorgia Ramponi , Niao He

The parameters for a Markov Decision Process (MDP) often cannot be specified exactly. Uncertain MDPs (UMDPs) capture this model ambiguity by defining sets which the parameters belong to. Minimax regret has been proposed as an objective for…

Artificial Intelligence · Computer Science 2023-02-14 Marc Rigter , Bruno Lacerda , Nick Hawes

We study infinite-horizon average-reward constrained Markov decision processes (CMDPs) under the unichain assumption and general policy parameterizations. Existing regret analyses for constrained reinforcement learning largely rely on…

Machine Learning · Computer Science 2026-02-10 Anirudh Satheesh , Vaneet Aggarwal

Model-based reinforcement learning (MBRL) agents typically learn world models by minimizing predictive loss. However, powerful RL optimizers inevitably exploit minor model inaccuracies, leading to simulator exploitation and a reality gap…

Machine Learning · Computer Science 2026-05-29 Christoph Dann , Yishay Mansour , Mehryar Mohri