Multi-stage genetic algorithm learning in game playing

We explore the concept of genetic structural expansion by employing a multi-stage chromosome coding scheme in a genetic algorithm (GA) based game-playing environment. Although structural expansion has been considered as a means of increasing diversity so as to benefit the GA optimization process in a changing world, it was seldom studied in the context of a multi-stage reinforced environment. This paper compares three chromosome coding schemes: monoploidy, triploidy (as a special case of polyploidy), and structural expansion, and discusses their impacts on multiple fuzzy-staged game-playing strategies. We show that when polyploid chromosomes are employed to cope with the changing environment in the domain of game-playing, the average learning result is apparently better than the learning curves in which only monoploidy is used