Genetic Local Search with Adaptive Crossover Probability and its Application to Maximum Diversity Problem

Abstract

This paper proposes a genetic local search (GLS) with crossover probability based on the distance between parent individuals. According to the distance between two selected parent individuals, the probability of crossover is decided. This probability is varied depending on the fitness values of parents and their child. By using adaptively varying the probability, the optimal distance between two parents for crossover is estimated. The proposed method is applied to the maximum diversity problem and compared with the previous GLS.