多群体并行遗传算法迁移策略的性能研究  

Empirically Investigation on Performances of Migration Policies of Multi-population Parallel Genetic Algorithms

作　　者：赖鑫生[1] 谭国律[1] 周玉林[1] 

机构地区：[1]上饶师范学院数学与计算机系,江西上饶334001

出　　处：《华东理工大学学报（自然科学版）》2007年第B06期51-53,99,共4页Journal of East China University of Science and Technology

摘　　要：研究了迁移策略对并行遗传算法性能的影响。研究了所有可能的4种组合：（1）选取最优个体进行迁移并在目的子群体中替换最差个体；（2）随机选取个体进行迁移并在目的子群体中替换最差个体；（3）选取最优个体进行迁移并在目的子群体中随机替换个体；（4）随机选取个体进行迁移并在目的子群体中随机替换个体。从累积量（Cumulants）的分析来看，似乎选取最优个体进行迁移并在目的子群体中随机替换个体的迁移策略能更好地兼顾局部搜索和全局搜索。通过对几个典型测试函数的测试，结果表明：选取最优个体进行迁移并在目的子群体中随机替换个体的迁移策略不比其他3种策略差，甚至在部分测试中要优于其他策略。这一结果可能有助于设计出性能更好的并行遗传算法。In this paper we empirically investigate how the policy used to select the migrants and the individuals to be replaced affects the performances of parallel evolutionary algorithms. The four possible combinations of random and fitness-based emigration and replacement of existing individuals are considered. From those analysis of cumulants, it seems that the best replacing individuals randomly could balance the exploits the information gathered about the problem with the genetic operators that explore new solution better, which is substantiated by the results of simulative experiments. We show that, on a variety of difficult multi-modal test functions, the best-replace-random policy does no worse than the others, and performs better on a subset of them. These results may help to design better-performance parallel evolutionary algorithms.

关 键 词：多种群 迁移策略 多样性 累积量 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]