Reduced Space and Faster Convergence in Imperfect-Information Games via Pruning

[edit]

Noam Brown, Tuomas Sandholm ;
Proceedings of the 34th International Conference on Machine Learning, PMLR 70:596-604, 2017.

Abstract

Iterative algorithms such as Counterfactual Regret Minimization (CFR) are the most popular way to solve large zero-sum imperfect-information games. In this paper we introduce Best-Response Pruning (BRP), an improvement to iterative algorithms such as CFR that allows poorly-performing actions to be temporarily pruned. We prove that when using CFR in zero-sum games, adding BRP will asymptotically prune any action that is not part of a best response to some Nash equilibrium. This leads to provably faster convergence and lower space requirements. Experiments show that BRP results in a factor of 7 reduction in space, and the reduction factor increases with game size.

Related Material