Reinforcement learning produces dominant strategies for the Iterated Prisoner's Dilemma

摘要

We present tournament results and several powerful strategies for the Iterated Prisoner'sudDilemma created using reinforcement learning techniques (evolutionary and particle swarmudalgorithms). These strategies are trained to perform well against a corpus of over 170 distinctudopponents, including many well-known and classic strategies. All the trained strategiesudwin standard tournaments against the total collection of other opponents. The trained strategiesudand one particular human made designed strategy are the top performers in noisy tournamentsudalso.