Genetic crossover strategy using an approximation concept

Abstract

This paper presents a crossover strategy which utilizes an approximation concept to steer the locations of progeny in genetic algorithm. The search domain is approximated by an n-dimensional second-order response surface in which each dimension corresponds to one design variable. Based on the assumption of second-order response, a quadratic curve is fitted through each design variable of the crossover pair. The value of each design variable in one of the progeny is then determined by the location where the derivative of the fitted surface vanishes. The use of such an approximation concept provides the capability of quickly moving progeny toward regions with improved fitness and can perform multi-dimensional search in parallel. Empirical results demonstrated that genetic algorithms with this crossover strategy could greatly accelerate the speed of obtaining optimal solutions.