作　　者： (赵新超）; (刘朝华）;

机构地区： 北京邮电大学理学院,北京100876

出　　处： 《集美大学学报（自然科学版）》 2017年第4期58-67,共10页

摘　　要： 为了充分利用记忆库内保存的有益历史信息,对解向量的微调进行差分变异操作的差分改进,提出了一种融入差分变异操作的变规模和声搜索(linearly decreasing harmony search,LDHS)算法,加强了算法的新路径探索性能。同时,为了平衡记忆库的多样性和收敛性,对记忆库的大小采取线性调整,并研究了记忆库大小对LDHS算法性能的影响。最后,将LDHS算法、基本的和声搜索(harmony search,HS)算法、3种改进的HS算法在CEC 2014的8个测试函数上分别进行不同维度独立运行30次实验,对比结果表明,LDHS算法能够更快地找到全局最优解,并具有较好的稳定性。 In order to make full use of the information of the harmony memory, a differential mutation operation was proposed to provid an improved direction for the fine tuning of the solution. A harmony search algorithm was also proposed with linearly decreasing population size-LDHS, which enhanced the exploration capability for the novel routes of the algorithm. At the same time, in order to balance the diversity and con- vergence of harmony memory, a linearly decreasing strategy was adopted to the population size of the harmony memory, and its effect was considered on the performance of the algorithm. Finally, the performance of the proposed algorithm was verified with the basic HS and other three improved HS variants on the 8 CEC 2014 benchmark functions in 30 independent runs on different dimensions. The experimental results showed that LDHS algorithm could find the even better solutions more efficiently, and also had better steadiness.

关 键 词： 和声搜索算法 变规模 差分变异 历史信息