English

A Black-box Approach for Non-stationary Multi-agent Reinforcement Learning

Machine Learning 2024-05-06 v2 Artificial Intelligence Computer Science and Game Theory Multiagent Systems Machine Learning

Abstract

We investigate learning the equilibria in non-stationary multi-agent systems and address the challenges that differentiate multi-agent learning from single-agent learning. Specifically, we focus on games with bandit feedback, where testing an equilibrium can result in substantial regret even when the gap to be tested is small, and the existence of multiple optimal solutions (equilibria) in stationary games poses extra challenges. To overcome these obstacles, we propose a versatile black-box approach applicable to a broad spectrum of problems, such as general-sum games, potential games, and Markov games, when equipped with appropriate learning and testing oracles for stationary environments. Our algorithms can achieve O~(Δ1/4T3/4)\widetilde{O}\left(\Delta^{1/4}T^{3/4}\right) regret when the degree of nonstationarity, as measured by total variation Δ\Delta, is known, and O~(Δ1/5T4/5)\widetilde{O}\left(\Delta^{1/5}T^{4/5}\right) regret when Δ\Delta is unknown, where TT is the number of rounds. Meanwhile, our algorithm inherits the favorable dependence on number of agents from the oracles. As a side contribution that may be independent of interest, we show how to test for various types of equilibria by a black-box reduction to single-agent learning, which includes Nash equilibria, correlated equilibria, and coarse correlated equilibria.

Keywords

Cite

@article{arxiv.2306.07465,
  title  = {A Black-box Approach for Non-stationary Multi-agent Reinforcement Learning},
  author = {Haozhe Jiang and Qiwen Cui and Zhihan Xiong and Maryam Fazel and Simon S. Du},
  journal= {arXiv preprint arXiv:2306.07465},
  year   = {2024}
}

Comments

26 Pages, 2 figures

R2 v1 2026-06-28T11:03:29.338Z