Genetic learning in strategic form games. (Q960458)

Summary: We analyze the learning behavior of a Simple Genetic Algorithm in symmetric \(3 \times 3\) Strategic-Form-Games. In cases of contests within one population and also between two populations the behavior of the SGA is compared with the behavior of the replicator dynamics and is analyzed with respect to equilibrium concepts in evolutionary game theory. Furthermore conservative non-adaptive strings are added to the population which lead to convergence to an equilibrium even in ``GA-deceptive'' games where the equilibrium can not be reached by GAs using only selection and crossover.