Fuzzy adaptation of crossover and mutation rates in genetic algorithms based on population performance

Abstract

A novel approach for the adaptive tuning of recombination rates of genetic algorithm through a fuzzy inference system is proposed. The method exploits a set of features assessing the status of the optimization process and determined on the basis of the fitness of a representative subset of the population. This features, at each generation, are fed to a fuzzy system for adjusting the mutation and crossover rates of the genetic algorithm. The method has been tested on classical problems that are often used in literature for assessing optimization algorithms. The achieved results show that this procedure improves the performance of the optimization process, by both speeding up the search, and avoiding the genetic algorithm to converge toward local minima.