和声搜索-分布估计混合算法求解多目标优化问题

针对和声搜索算法不能很好地求解多目标优化问题的缺陷,提出一种多目标和声搜索—分布估计混合算法(MHS-EDA)。该算法一方面利用分布估计的采样操作对和声记忆库内进行搜索,拓宽了和声记忆库内空间;另一方面对和声记忆库外进行外部档案搜索,实现群体间信息交换,从而提高了多目标和声算法的全局搜索能力。数值实验选取六个常用测试函数,并与多目标遗传算法、多目标分布估计算法、多目标和声搜索算法进行比较,测试结果表明提出的混合算法能够有效地解决多目标优化问题。

Hybrid harmony search and estimation of distribution algorithm formulti-objective optimization problems

For solving the problems of basic harmony search algorithm can't be well used for multi-objective optimization, this paper proposed a kind of multi-objective hybrid harmony search and estimation of distribution(MHS-EDA). In the harmony memory, sampled every variable by estimation of distribution, so it would widen the space of harmony memory. Out of the harmony memory, every variable would be searched by external repository, so it could exchange information between the populations during the proceeding of evolutionary and enhance the global searching ability of harmony search algorithm. Numerical experiments compare with multi-objective genetic algorithm, multi-objective estimation of distribution algorithm and multi-objective harmony search algorithm on six benchmark problems. The results show that the proposed algorithm can be effective to solve multi-objective optimization problems.