Memory Size, Learning and Dynamic Preferential Selection in the Spatial Prisoner's Dilemma Game

The prisoner's dilemma is widely used in the research of evolution of cooperation. This work is based on the assumption that players use non-discriminative strategies within their neighborhoods. The paper is also assumed that players have different memory sizes to store the previous interaction histories. Two different learning rules, copy-best-p layer and copy-best-strategy, are considered in this paper. In each round, a player can use either rule to select the appropriate strategy from his neighbors as his own strategy for the next round. The player uses preferential selection rule to select a neighbor to learn from. By the use of MC (Monte Carlo) simulation, research results are obtained as follows: 1) the preferential selection rule considerably improves the cooperation level in heterogeneous networks while it inhibits the emergence of cooperation in homogeneous regular network; 2) different learning rules and memory sizes significantly affect the evolution of cooperation in all types of network, especially in homogeneous network.