Model-Based Reinforcement Learning for Offline Zero-Sum Markov Games

This paper makes progress toward learning Nash equilibria in two-player, zero-sum Markov games from offline data. Specifically, consider a γ-discounted, infinite-horizon Markov game with S states, in which the max-player has A actions and the min-player has B actions. We propose a pessimistic model–based algorithm with Bernstein-style lower confidence bounds—called the value iteration with lower confidence bounds for zero-sum Markov games—that provably finds an ε-approximate Nash equilibrium with a sample complexity no larger than $\frac{C_{\mathsf{\text{clipped}}}^{\star }S(A+B)}{{(1-\gamma )}^3{\epsilon }^2}$ (up to some log factor). Here, $C_{\mathsf{\text{clipped}}}^{\star }$ is some unilateral clipped concentrability coefficient that reflects the coverage and distribution shift of the available data (vis-à-vis the target data), and the target accuracy ε can be any value within $\left(0,\frac{1}{1-\gamma }\right]$. Our sample complexity bound strengthens prior art by a factor of $\min \{A,B\}$, achieving minimax optimality for a broad regime of interest. An appealing feature of our result lies in its algorithmic simplicity, which reveals the unnecessity of variance reduction and sample splitting in achieving sample optimality.

Funding: Y. Yan is supported in part by the Charlotte Elizabeth Procter Honorific Fellowship from Princeton University and the Norbert Wiener Postdoctoral Fellowship from MIT. Y. Chen is supported in part by the Alfred P. Sloan Research Fellowship, the Google Research Scholar Award, the Air Force Office of Scientific Research [Grant FA9550-22-1-0198], the Office of Naval Research [Grant N00014-22-1-2354], and the National Science Foundation [Grants CCF-2221009, CCF-1907661, IIS-2218713, DMS-2014279, and IIS-2218773]. J. Fan is supported in part by the National Science Foundation [Grants DMS-1712591, DMS-2052926, DMS-2053832, and DMS-2210833] and Office of Naval Research [Grant N00014-22-1-2340].

Supplemental Material: The online appendix is available at https://doi.org/10.1287/opre.2022.0342.