On a self-organizing property of genetic algorithms

The dynamics of a genetic algorithm (GA) is investigated from the viewpoint of self-organization. The dynamics of the simple GA is modeled by difference equations. The equations are examined by computer simulation, and it is shown that the GA selects a genotype robust to a mutation. This property becomes remarkable when a crossover operator is introduced. The GA is applied to a simple optimization problem with redundant genetic coding. Results of computer simulations show that the GA selects a genetic representation that is robust to reduction of fitness value by the mutation and the crossover as well