Function optimization using an adaptive crossover operator based on locality

Abstract

This paper proposes a new crossover operator in genetic algorithms for function optimization. The proposed Adaptive Crossover Operator (ACO) restricts the crossover range by using a bias_value. The bias_value is computed by the fitness function value, the performance ratio, and the number of generations. This ACO can reduce the computational complexity of obtaining the global optimum. By maintaining diversity in the population and sustaining the balance between exploration and exploitation, it can also prevent the genetic algorithm from getting stuck at a local optimum. As the generations progress, the ACO achieves a local improvement on a part of a chromosome which is restricted by the bias_value. The experiment demonstrates that a genetic algorithm using the ACO performs better than a genetic algorithm using a standard crossover in optimizing several functions.