An adaptive method of Hungarian mating schemes in genetic algorithms

Abstract

Mating scheme is one of the key operations in genetic algorithms. In this paper, we propose an adaptive mating method combining Hungarian mating schemes that have been previously suggested. Hungarian mating schemes include i) minimizing the sum of matching distances, ii) maximizing the sum, and iii) random matching. Our adaptive mating method selects one of the schemes with voting. Every matched pair of individuals has the right to vote for the mating scheme of the next generation. Its preference is closely related to the ratio of distance between parents over distance between parent and offspring.We apply the proposed method to well-known combinatorial optimization problems, the traveling salesman problem and the graph bisection problem. The proposed adaptive method showed better performance than any single Hungarian mating scheme and the nonadaptive hybrid scheme presented in previous work.