文献类型：期刊文献

中文题名：基于败者组与混合编码策略的NSGA-Ⅱ改进算法

英文题名：Improved NSGA-Ⅱ Algorithm Based on Loser Group and Hybrid Coding Strategy

作者：刘鑫平[1];顾春华[1];罗飞[1];丁炜超[1]

机构：[1]华东理工大学信息科学与工程学院

年份：2019

卷号：46

期号：10

起止页码：222

中文期刊名：计算机科学

外文期刊名：Computer Science

收录：CSTPCD;;北大核心:【北大核心2017】；CSCD:【CSCD_E2019_2020】；

基金：国家自然科学基金项目(61472139)资助

语种：中文

中文关键词：多目标进化算法;NSGA-II;败者组;循环拥挤系数排序;混合编码

外文关键词：Multi-objective evolutionary algorithm;NSGA-II;Loser group;Cyclic congestion coefficient ranking;Hybrid coding

摘要：在精英选择中NSGA-II的拥挤系数算子对局部拥挤区域的分布性优化效果不佳,并且会使某些更接近Pareto最优解集的个体被淘汰。针对拥挤系数算子存在优秀个体不被保留的缺陷,提出了一种基于败者组与混合编码策略的改进算法(LGHC-NSGA-II)。参照棋类比赛中的双败淘汰制,构建了败者组外部归档集,在迭代结束后将归档集与末代父代种群合并,并采用循环拥挤系数排序策略优化分布性。同时,针对传统编码方式在全局或局部空间上搜索能力较差的缺陷,提出了一种混合编码策略,有效地提高了算法的收敛性。基于ZDT系列问题上的测试结果表明,改进算法与8种多目标进化算法相比,在算法的收敛性、分布性与鲁棒性上均具有较高的优越性。
The congestion coefficient operator of NSGA-II in elite selection can not optimize the distribution of local congestion area effectively,and some individuals closer to Pareto optimal solution set will be eliminated.An improved algorithm based on loser group and hybrid encoding strategy (LGHC-NSGA-II) was proposed to overcome the shortcoming that excellent individuals are not retained in the congestion coefficient operator.Referring to the double-losing elimination system in chess games,an external archive set of loser group is constructed.After the iteration,the archive set is merged with the last generation parent population,and the distribution coefficient is optimized by cyclic congestion coefficient ranking strategy.At the same time,a hybrid coding strategy was proposed to overcome the shortcomings of traditional coding methods in global or local space,which effectively improves the convergence of the algorithm.The test results on ZDT series problems show that the improved algorithm is superior to eight multi-objective evolutionary algorithms in convergence,distribution and robustness.