Uncoupled Bandit Learning towards Rationalizability: Benchmarks, Barriers, and Algorithms [article]

Jibang Wu, Haifeng Xu, Fan Yao
2023 arXiv   pre-print
Under the uncoupled learning setup, the last-iterate convergence guarantee towards Nash equilibrium is shown to be impossible in many games. This work studies the last-iterate convergence guarantee in general games toward rationalizability, a key solution concept in epistemic game theory that relaxes the stringent belief assumptions in both Nash and correlated equilibrium. This learning task naturally generalizes best arm identification problems, due to the intrinsic connections between
more » ... izable action profiles and the elimination of iteratively dominated actions. Despite a seemingly simple task, our first main result is a surprisingly negative one; that is, a large and natural class of no regret algorithms, including the entire family of Dual Averaging algorithms, provably take exponentially many rounds to reach rationalizability. Moreover, algorithms with the stronger no swap regret also suffer similar exponential inefficiency. To overcome these barriers, we develop a new algorithm that adjusts Exp3 with Diminishing Historical rewards (termed Exp3-DH); Exp3-DH gradually forgets history at carefully tailored rates. We prove that when all agents run Exp3-DH (a.k.a., self-play in multi-agent learning), all iteratively dominated actions can be eliminated within polynomially many rounds. Our experimental results further demonstrate the efficiency of Exp3-DH, and that state-of-the-art bandit algorithms, even those developed specifically for learning in games, fail to reach rationalizability efficiently.
arXiv:2111.05486v3 fatcat:qtglhexxtbc6hlqwggnnjogwwu