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

arXiv:2306.07465v2 Announce Type: replace
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 …

abstract agent arxiv box challenges cs.ai cs.gt cs.lg cs.ma equilibria equilibrium feedback focus games gap multi-agent multi-agent learning reinforcement reinforcement learning small stat.ml systems testing type