基于岛屿群体模型的多目标演化算法研究  

Research on Multi-objective Evolutionary Algorithm Based on Island Model

作　　者：赵凤强[1] 徐毅[2] 李广强[3] 

机构地区：[1]大连民族学院机电信息工程学院,大连116600 [2]大连理工大学计算机科学与工程系,大连116023 [3]大连海事大学信息科学技术学院,大连116026

出　　处：《计算机科学》2010年第12期190-192,205,共4页Computer Science

基　　金：国家自然科学基金(60674061)资助

摘　　要：近年来,基于Pareto最优概念的多目标演化算法成为演化计算的研究热点,并已在工程领域中得到了广泛应用。在多目标演化算法NSGA-II基础上,给出了并行非劣分层多目标演化算法(PNSMEA)。该算法引入了粗粒度岛屿模型,整个群体被划分成若干个子群体,每个子群体单独演化计算。子群体在进化过程中,每隔一定进化代数交换集合中的个体,以保证各子群体中个体的多样性,提高多目标问题非劣最优域搜索的广度。引入算术交叉算子,以克服NSGA-II中SBX(Simulated Binary Crossover)交叉算子搜索能力较弱的缺点。试验结果表明,PNSMEA算法不仅可改善NSGA-II算法的搜索孤立区域困难和早收敛的问题,而且所获得的Pareto解集具有更好的分布性。Recently the reaserch on multi-objective evolutionary algorithms based on Pareto optimization concept has become a research hotspot.And it has been widely applied in engineering fields.This paper presented a parallel nondominated sorting genetic multi-objective evolutionary algorithm（PNSMEA） based on NSGA-II.PNSMEA adopes island model and the population is divided into several sub-populations that evolve separately.The sub-populations migrate good individules each other at intervals of some generations,which can keep individules＇ diversity and broad the search domain of each sub-population.PNSMEA adopes arithmetic crossover operator to overcome the weak search capability of SBX operator used by NSGA-II.The test results show that PNSMEA can not only improve the premature problem as well as the search capability in the isolated regions of NSGA-II but also contribute to obtaining the Pareto solution sets with better distribution.

关 键 词：多目标演化算法 NSGA-II PARETO解集 岛屿模型 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]