The Power of Exploiter: Provable Multi-Agent RL in Large State Spaces

Chi Jin, Qinghua Liu, Tiancheng Yu
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:10251-10279, 2022.

Abstract

Modern reinforcement learning (RL) commonly engages practical problems with large state spaces, where function approximation must be deployed to approximate either the value function or the policy. While recent progresses in RL theory address a rich set of RL problems with general function approximation, such successes are mostly restricted to the single-agent setting. It remains elusive how to extend these results to multi-agent RL, especially in the face of new game-theoretical challenges. This paper considers two-player zero-sum Markov Games (MGs). We propose a new algorithm that can provably find the Nash equilibrium policy using a polynomial number of samples, for any MG with low multi-agent Bellman-Eluder dimension—a new complexity measure adapted from its single-agent version (Jin et al., 2021). A key component of our new algorithm is the exploiter, which facilitates the learning of the main player by deliberately exploiting her weakness. Our theoretical framework is generic, which applies to a wide range of models including but not limited to tabular MGs, MGs with linear or kernel function approximation, and MGs with rich observations.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-jin22c, title = {The Power of Exploiter: Provable Multi-Agent {RL} in Large State Spaces}, author = {Jin, Chi and Liu, Qinghua and Yu, Tiancheng}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {10251--10279}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/jin22c/jin22c.pdf}, url = {https://proceedings.mlr.press/v162/jin22c.html}, abstract = {Modern reinforcement learning (RL) commonly engages practical problems with large state spaces, where function approximation must be deployed to approximate either the value function or the policy. While recent progresses in RL theory address a rich set of RL problems with general function approximation, such successes are mostly restricted to the single-agent setting. It remains elusive how to extend these results to multi-agent RL, especially in the face of new game-theoretical challenges. This paper considers two-player zero-sum Markov Games (MGs). We propose a new algorithm that can provably find the Nash equilibrium policy using a polynomial number of samples, for any MG with low multi-agent Bellman-Eluder dimension—a new complexity measure adapted from its single-agent version (Jin et al., 2021). A key component of our new algorithm is the exploiter, which facilitates the learning of the main player by deliberately exploiting her weakness. Our theoretical framework is generic, which applies to a wide range of models including but not limited to tabular MGs, MGs with linear or kernel function approximation, and MGs with rich observations.} }
Endnote
%0 Conference Paper %T The Power of Exploiter: Provable Multi-Agent RL in Large State Spaces %A Chi Jin %A Qinghua Liu %A Tiancheng Yu %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-jin22c %I PMLR %P 10251--10279 %U https://proceedings.mlr.press/v162/jin22c.html %V 162 %X Modern reinforcement learning (RL) commonly engages practical problems with large state spaces, where function approximation must be deployed to approximate either the value function or the policy. While recent progresses in RL theory address a rich set of RL problems with general function approximation, such successes are mostly restricted to the single-agent setting. It remains elusive how to extend these results to multi-agent RL, especially in the face of new game-theoretical challenges. This paper considers two-player zero-sum Markov Games (MGs). We propose a new algorithm that can provably find the Nash equilibrium policy using a polynomial number of samples, for any MG with low multi-agent Bellman-Eluder dimension—a new complexity measure adapted from its single-agent version (Jin et al., 2021). A key component of our new algorithm is the exploiter, which facilitates the learning of the main player by deliberately exploiting her weakness. Our theoretical framework is generic, which applies to a wide range of models including but not limited to tabular MGs, MGs with linear or kernel function approximation, and MGs with rich observations.
APA
Jin, C., Liu, Q. & Yu, T.. (2022). The Power of Exploiter: Provable Multi-Agent RL in Large State Spaces. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:10251-10279 Available from https://proceedings.mlr.press/v162/jin22c.html.

Related Material