A self-adaptive genetic algorithm for function optimization

Genetic algorithms (GA's) have some control pa rameters such as the probability of bit mutation or the probability of crossover. These are nornially given a priori by the user (programmer) of the algorithm. There exists a wide variety of values for control parameters and it is difficult to find the best choice of these values in order to optimize the be haviour of a particular GA. We introduce a self adaptive GA (SAGA) with its control parameters encoded in the genome of the individuals of the population. This algorithm is used to optimize a set of twenty functions from R2 to R and its behaviour is compared with the one resulting from the execution of a traditional GA varying its control parameter values. We obtain a set. of measurements which demonstrate statistically that SAGA yields a set of results which compare favourably with the same results mean values from an extensive set of runs of traditional GA (TGA).