arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2007.07461cs.LGcs.GTcs.MAmath.OCstat.ML
收藏

零和马尔可夫博弈中基于模型的多智能体强化学习的近最优样本复杂度

Model-Based Multi-Agent RL in Zero-Sum Markov Games with Near-Optimal Sample Complexity

University of Maryland, College Park(马里兰大学帕克分校);Harvard University(哈佛大学);University of Illinois at Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校);University of California, Los Angeles(加利福尼亚大学洛杉矶分校)

Kaiqing Zhang, Sham M. Kakade, Tamer Başar, Lin F. Yang

AI总结:

本文研究双人零和马尔可夫博弈中基于模型的多智能体强化学习,证明其样本复杂度为$\ ilde O(|S||A||B|(1-γ)^{-3}ε^{-2})$,并证明该界在奖励无关设置下极小极大最优,揭示了其能力与局限性的基本权衡。

摘要:

基于模型的强化学习(RL)利用经验模型寻找最优策略,长期以来被认为是强化学习的基石之一。它特别适用于多智能体强化学习(MARL),因为它自然地将学习阶段与规划阶段解耦,并避免了当所有智能体同时使用样本改进策略时出现的非平稳性问题。尽管这种方法直观且被广泛使用,基于模型的MARL算法的样本复杂度尚未得到充分研究。本文的目标是解决关于其样本复杂度的基本问题。我们研究了可以说是最基本的MARL设置:双人折扣零和马尔可夫博弈,且只能访问生成模型。我们证明,基于模型的MARL在寻找纳什均衡(NE)值达到某个ε误差时,样本复杂度为$\ ilde O(|S||A||B|(1-γ)^{-3}ε^{-2})$,并在具有平滑规划预言机的情况下找到ε-NE策略,其中γ是折扣因子,S、A、B分别表示状态空间和两个智能体的动作空间。我们进一步证明,如果算法是奖励无关的(即算法在查询状态转移样本时没有奖励知识),通过建立匹配的下界,这样的样本界是极小极大最优的(最多相差对数因子)。这与通常的奖励感知设置形成对比,后者的下界为$\ ildeΩ(|S|(|A|+|B|)(1-γ)^{-3}ε^{-2})$,在这种设置下,基于模型的方法仅在|A|、|B|依赖性上存在差距,因此是近最优的。我们的结果不仅证明了这种基本基于模型的方法在MARL中的样本效率,还详细阐述了其能力(容易处理更具挑战性的奖励无关情况)与局限性(在|A|、|B|上适应性较差且次优)之间的基本权衡,这种权衡尤其出现在多智能体情境中。

英文摘要:

Model-based reinforcement learning (RL), which finds an optimal policy using an empirical model, has long been recognized as one of the corner stones of RL. It is especially suitable for multi-agent RL (MARL), as it naturally decouples the learning and the planning phases, and avoids the non-stationarity problem when all agents are improving their policies simultaneously using samples. Though intuitive and widely-used, the sample complexity of model-based MARL algorithms has not been fully investigated. In this paper, our goal is to address the fundamental question about its sample complexity. We study arguably the most basic MARL setting: two-player discounted zero-sum Markov games, given only access to a generative model. We show that model-based MARL achieves a sample complexity of $\tilde O(|S||A||B|(1-γ)^{-3}ε^{-2})$ for finding the Nash equilibrium (NE) value up to some $ε$ error, and the $ε$-NE policies with a smooth planning oracle, where $γ$ is the discount factor, and $S,A,B$ denote the state space, and the action spaces for the two agents. We further show that such a sample bound is minimax-optimal (up to logarithmic factors) if the algorithm is reward-agnostic, where the algorithm queries state transition samples without reward knowledge, by establishing a matching lower bound. This is in contrast to the usual reward-aware setting, with a $\tildeΩ(|S|(|A|+|B|)(1-γ)^{-3}ε^{-2})$ lower bound, where this model-based approach is near-optimal with only a gap on the $|A|,|B|$ dependence. Our results not only demonstrate the sample-efficiency of this basic model-based approach in MARL, but also elaborate on the fundamental tradeoff between its power (easily handling the more challenging reward-agnostic case) and limitation (less adaptive and suboptimal in $|A|,|B|$), particularly arises in the multi-agent context.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑