Reinforcement learning with restrictions on the action set

Mario Bravo, Mathieu Faure

Consider a 2-player normal-form game repeated over time. We introduce an adaptive learning procedure, where the players only observe their own realized payoff at each stage. We ass…