A differential evolution algorithm to develop strategies for the iterated prisoner’s dilemma

Summarization: This paper presents the application of the Differential Evolution (DE) algorithm in the most known dilemma in the field of Game Theory, the Prisoner’s Dilemma (PD) that simulates the selfish behavior between rational individuals. This study investigates the suitability of the DE to evolve strategies for the Iterated Prisoner’s Dilemma (IPD), so that each individual in the population represents a complete playing strategy. Two different approaches are presented: a classic DE algorithm and a DE approach with memory. Their results are compared with several benchmark strategies. In addition, the Particle Swarm Optimization (PSO) and the Artificial Bee Colony (ABC) that have been implemented in the same framework are compared with the DE approaches. Overall, the strategies developed by DE outperform all the others. Also, it has been observed over iterations that when the DE algorithm is used the player manages to learn his opponent, therefore, DE converges with a quick and efficient manner.Παρουσιάστηκε στο: 3rd International Conference on Machine Learning, Optimization, and Big Dat