PDF全文下载:2016040457
梁存利
(西藏民族大学 教育学院,陕西 咸阳712082)
摘要: 提出了一个基于Taguchi方法的双种群混合粒子群算法。首先新算法设计了与种群规模和维数相关的可行解空间分割方法,使得初始种群的个体能尽量均匀地分布在可行解空间内,为算法的搜索提供了一个良好的初始平台;其次设计的Taguchi学习策略不仅能有效地整合种群中其它粒子历史最好位置的信息,而且能充分地交换了两个种群当前最优解的信息,从而很好地维持了种群的多样性,有效避免算法陷入早熟收敛。 通过14个标准的测试函数验证了新算法的有效性。
关键词:粒子群算法;遗传算法;算术杂交;Taguchi; 约束优化
中图分类号:TP 18文献标志码:A
A Hybrid Bi-Population Particle Swarm Optimizer Combined with Taguchi MethodLIANG Cunli
(College of Education, Xizang Minzu University, Xianyang 712082, China)
Abstract: A hybrid bi-population particle swarm optimizer combined with Taguchi method is put forward. According to population size and problem dimension,the feasible solution space is divided to scatter an initial population of points uniformly in it. The Taguchi learning strategy can not only integrate all other particle′s historical best information effectively but also combine the information of the best positions discovered by the two populations, which keeps the diversity of population and avoids premature convergence. Experiments are conducted on fourteen multimodal test functions.The results demonstrate good performance of the proposed algorithm.
Key words: particle swarm optimizer; genetic algorithm;arithmetic crossover;taguchi; constrained optimization
收稿日期: 20150709
基金项目: 西藏自治区自然科学基金项目(2015ZR1325).
作者简介: 梁存利(1969—),男,副教授.
文章编号:16726987(2016)04045708; DOI: 10.16351/j.16726987.2016.04.020