Learn Your Opponent's Strategy (in Polynomial Time)!

Agents that interact in a distributed environment might increase their utility by behaving optimally given the strategies of the other agents. To do so, agents need to learn about those with whom they share the same world. This paper examines interactions among agents from a game theoretic perspective. In this context, learning has been assumed as a means to reach equilibrium. We analyze the complexity of this learning process. We start with a restricted two--agent model, in which agents are represented by finite automata, and one of the agents plays a fixed strategy. We show that even with this restrictions, the learning process may be exponential in time. We then suggest a criterion of simplicity, that induces a class of automata that are learnable in polynomial time