Provable Memory Efficient Self-Play Algorithm for Model-free Reinforcement Learning

TL;DR

This paper introduces ME-Nash-QL, a memory-efficient, model-free self-play algorithm for two-player zero-sum Markov games that improves space, sample, and burn-in costs while maintaining Markov policies.

Contribution

The paper proposes ME-Nash-QL, the first algorithm with provable memory efficiency and improved complexity bounds for model-free MARL in two-player zero-sum Markov games.

Findings

Achieves space complexity $O(SABH)$ for $\e$-approximate Nash policy.

Reduces sample complexity to $ ilde{O}(H^4SAB/\e^2)$ for long horizons.

Lowers burn-in cost to $O(SAB ext{poly}(H))$ compared to previous methods.

Abstract

The thriving field of multi-agent reinforcement learning (MARL) studies how a group of interacting agents make decisions autonomously in a shared dynamic environment. Existing theoretical studies in this area suffer from at least two of the following obstacles: memory inefficiency, the heavy dependence of sample complexity on the long horizon and the large state space, the high computational complexity, non-Markov policy, non-Nash policy, and high burn-in cost. In this work, we take a step towards settling this problem by designing a model-free self-play algorithm \emph{Memory-Efficient Nash Q-Learning (ME-Nash-QL)} for two-player zero-sum Markov games, which is a specific setting of MARL. ME-Nash-QL is proven to enjoy the following merits. First, it can output an $\varepsilon$-approximate Nash policy with space complexity $O(SABH)$ and sample complexity $\widetilde{O}(H^4SAB/\varepsilon^2)$, where $S$ is the number of states, $\{A, B\}$ is the number of actions for two players, and $H$ is the horizon length. It outperforms existing algorithms in terms of space complexity for tabular cases, and in terms of sample complexity for long horizons, i.e., when $\min\{A, B\}\ll H^2$. Second, ME-Nash-QL achieves the lowest computational complexity $O(T\mathrm{poly}(AB))$ while preserving Markov policies, where $T$ is the number of samples. Third, ME-Nash-QL also achieves the best burn-in cost $O(SAB\,\mathrm{poly}(H))$, whereas previous algorithms have a burn-in cost of at least $O(S^3 AB\,\mathrm{poly}(H))$ to attain the same level of sample complexity with ours.