Genetic Algorithm Using an Adaptive Mutation Operator for Numerical Optimization Functions

Abstract

Simple genetic algorithms (GAs) applies only one mutation operator to produce next population. Four mutation operators have been suggested in GAs. Specifying which mutation operator should be employed is very difficult but not impossible and is normally done by empirical results. In this paper, new adaptive approach is proposed to solve the problem. This scheme uses more than one mutation operator to produce next generation. It modifies the selection mutation ratio of each mutation operator based on the global behaviour of the population for each generation. The main idea of the adaptive algorithm is to alter the operators’ probabilities based on the feedback information of population for each generation. This technique is conducted on a fifteen benchmark optimization problems. The result show that the adaptive mutation accomplishes equally well for all of the test functions.