核模糊聚类划分子种群的双种群遗传算法  

Multi-population genetic algorithm of two populations for dividing subpopulations by kernel fuzzy clustering

作　　者：孙雨萌 柏丽娜 张旭秀[1] SUN Yumeng;BAI Lina;ZHANG Xuxiu(School of Electrical Information Engineering,Dalian Jiaotong University,Dalian 116028,China)

机构地区：[1]大连交通大学电气信息工程学院,辽宁大连116028

出　　处：《大连工业大学学报》2021年第1期67-73,共7页Journal of Dalian Polytechnic University

基　　金：国家科技支撑计划项目(2015BAF20B02);国家自然科学基金项目(61471080,61201419).

摘　　要：并行双种群遗传算法在一定程度上避免了单一机制遗传算法易出现“早熟”的现象,但在其迭代进化后期存在种群同质化严重的缺陷。针对这一问题,在进行种群划分时引入核模糊聚类算法,将个体适应度值作为双种群聚类划分的约束条件,并针对划分所得双种群,提出两种改进的自适应交叉及变异策略,分别侧重遗传算法中局部搜索能力和全局探索能力。通过典型测试函数进行验证,对比标准双种群遗传算法(2PMGA)及自适应双种群遗传算法(A-2PMGA)。实验表明,所提出的核模糊聚类划分子种群的双种群遗传算法有效地解决了种群同质化的问题,避免子种群陷入同一局部最优值。Although parallel two-population genetic algorithm avoids the premature phenomenon which is easy to occur in a single mechanism genetic algorithm to some extent,there is a serious defect in homogenization between populations in the late iterative evolution.Aiming to solve the above problems,a kernel fuzzy clustering algorithm in population division is introduced,and used the individual fitness value as the constraint condition of the two-population clustering.Two improved adaptive crossovers and mutations were proposed for the divided two populations,focusing on the local search ability and global exploration ability in the genetic algorithm respectively.The typical test function was used to verify and compare the standard Multi-population genetic algorithm of two populations(2PMGA)and the adaptive Multi-population genetic algorithm of two populations(A-2PMGA).The experiment shows that the improved two-population genetic algorithm based on the kernel fuzzy clustering can effectively solve the problem of the population homogenization and prevent subpopulations from falling into the same local optimal value.

关 键 词：双种群遗传算法 核模糊聚类分析 自适应交叉 自适应多位变异 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]