囚人のジレンマの最適戦略の進化的学習手法による獲得

概要

論文の詳細を見る
The prisoner's dilemma is well known in the field of game theory. In this paper, we searched for the strategy that gives a high gain in the repeated prisoner's dilemma. We first express the prisoner's dilemma as the Mealy machine (a finite automaton that the output is decided according to its present states and inputs), and search its structure by the genetic algorithm. Experiments are carried out to examine how the proposed technique is effective. In the experiments, "the direct coding method" is compared with the Mealy machine method. As a result, a higher gain is obtained with the Mealy machine method than with the simple direct coding method. In addition, it is confirmed that the Mealy machine is applicable no matter how many times the prisoner's dilemma is repeated.