文献类型：期刊文献

中文题名：基于动态扩张角的广义Pareto支配优化算法

英文题名：Generalizaed Pareto Domination Optimization Algorithm Based on Dynamic Expansion Angle

作者：郝新东[1];祁荣宾[1]

机构：[1]华东理工大学信息科学与工程学院,上海200237

年份：2018

卷号：44

期号：4

起止页码：609

中文期刊名：华东理工大学学报（自然科学版）

外文期刊名：Journal of East China University of Science and Technology

收录：CSTPCD;;Scopus;北大核心:【北大核心2017】；CSCD:【CSCD_E2017_2018】；

基金：上海市自然科学基金(15ZR1408900;14ZR1410500)

语种：中文

中文关键词：高维多目标优化;动态扩张角;广义支配;选择压

外文关键词：high-dimensional multi-objective optimization;dynamic expansion angle;generalized domination;selective pressure

摘要：NSGA-Ⅱ算法在处理高维多目标问题时解集的区分度变得很差,对此,有学者提出了基于扩张角的广义Pareto支配优化算法(GPO-NSGA-Ⅱ),即通过改变扩张角来调整解的支配区域,从而调整解集的区分度,进化过程中扩张角保持恒定。本文在GPO-NSGA-Ⅱ算法的基础上提出了随着种群进化扩张角动态改变的广义Pareto支配优化算法(DGPO-NSGA-Ⅱ),通过动态调整种群进化过程中的扩张角来影响种群进化的选择压。扩张角的动态调整采用线性减小方式,即随着种群的进化将扩张角从初始扩张角线性减小为0。为保证获得一个较好的初始扩张角区间,对种群进化的不同扩张角进行了大量对比实验。将该算法与GPO-NSGA-Ⅱ、NSGA-Ⅱ在测试函数上进行对比实验,结果表明该算法能以更高的精度更快地收敛到理论前沿,个体分布也更均匀。
The NSGA-Ⅱ algorithm has poor discrimination on the solution set during dealing with high dimension multi objective evolutionary problems. Aiming at the above shortcoming, a generalized Pareto domination optimization algorithm based on the expansion angle （GPO NSGA-Ⅱ ） was proposed, whose feature is to change the expansion angle so as to adjust the dominance area of solutions and raise the degree of discriminability. In the evolutionary process of the GPO NSGA-Ⅱ algorithm, the algorithm＇s expansion angle will remain constant. In this paper, we propose a dynamic generalized Pareto domination optimization algorithm, DGPO NSGA-Ⅱ. By dynamically adjusting the expansion angle in the population evolution process, the selection pressure of the population evolution may be affected. The dynamic adjustment of the expansion angle is linearly reduced, that is, the expansion angle is decreased linearly from the initial expansion angle to 0 as the population evolves. In order to ensure a better initial expansion angle interval, a large number of comparative experiments are carried out on the different expansion angles of population evolution. Finally, by comparing with GPO NSGA-Ⅱ and NSGA-Ⅱ in the test function, the proposed algorithm can converge the individual is more uniform.