English

Near-optimal Policy Optimization Algorithms for Learning Adversarial Linear Mixture MDPs

Machine Learning 2022-04-21 v2 Optimization and Control Machine Learning

Abstract

Learning Markov decision processes (MDPs) in the presence of the adversary is a challenging problem in reinforcement learning (RL). In this paper, we study RL in episodic MDPs with adversarial reward and full information feedback, where the unknown transition probability function is a linear function of a given feature mapping, and the reward function can change arbitrarily episode by episode. We propose an optimistic policy optimization algorithm POWERS and show that it can achieve O~(dHT)\tilde{O}(dH\sqrt{T}) regret, where HH is the length of the episode, TT is the number of interactions with the MDP, and dd is the dimension of the feature mapping. Furthermore, we also prove a matching lower bound of Ω~(dHT)\tilde{\Omega}(dH\sqrt{T}) up to logarithmic factors. Our key technical contributions are two-fold: (1) a new value function estimator based on importance weighting; and (2) a tighter confidence set for the transition kernel. They together lead to the nearly minimax optimal regret.

Keywords

Cite

@article{arxiv.2102.08940,
  title  = {Near-optimal Policy Optimization Algorithms for Learning Adversarial Linear Mixture MDPs},
  author = {Jiafan He and Dongruo Zhou and Quanquan Gu},
  journal= {arXiv preprint arXiv:2102.08940},
  year   = {2022}
}

Comments

22 pages, 1 figure. In AISTATS 2022

R2 v1 2026-06-23T23:15:39.559Z