Adaptive crossover operator based on locality and convergence

Abstract

In this paper, we propose an adaptive crossover operator (ACO) for function optimization. ACO performs local improvement by restricting the crossover range in adaptive way. This approach is based on bias value which restrict the location of crossover point. Bias value is computed by the fitness function value performance ratio, and the number of generations. As generations progress, the portion of chromosome to apply ACO becomes much smaller. ACO scheme can reduce the computation complexity and escape from getting stuck local optimum and also by maintaining diversity if can maintain the balance between exploration and exploitation. Several experiments have been carried our to compare the performance of adaptive scheme and standard scheme. Compared to simple GA, the proposed method is faster and more accurate in finding global optimum.