On the Convergence of Reinforcement Learning

This paper examines the convergence of payoffs and strategies in Erev and Roth`s model of reinforcement learning. When all players use this rule it eliminates iteratively dominated strategies and in two-person constant-sum games average payoffs converge to the value of the game. Strategies converge in constant-sum games with unique equilibria if they are pure or in 2 × 2 games also if they are mixed. The long-run behaviour of the learning rule is governed by equations related to Maynard Smith`s version of the replicator dynamic. Properties of the learning rule against general opponents are also studied. In particular it is shown that it guarantees that the lim sup of a player`s average payoffs is at least his minmax payoff.