Decentralized learning in two-player zero-sum games: A LR-I lagging anchor algorithm

This paper presents a L
_R-I
lagging anchor algorithm that combines a lagging anchor method to the L
_R-I
learning algorithm. We prove that this decentralized learning algorithm converges in strategies to a Nash equilibrium in two-player, zero-sum, two-action matrix games, while only needing knowledge of their own action and reward.