Computer Engineering and Applications ›› 2009, Vol. 45 ›› Issue (10): 49-53.DOI: 10.3778/j.issn.1002-8331.2009.10.015

• 研究、探讨 • Previous Articles     Next Articles

Increase diversity of solutions of MOEAs—∞-norm based stepwise method

ZENG Ying-lan,ZHENG Jin-hua,LUO Biao   

  1. Institute of Information Engineering,Xiangtan University,Xiangtan,Hunan 411105,China
  • Received:2008-05-28 Revised:2008-08-04 Online:2009-04-01 Published:2009-04-01
  • Contact: ZENG Ying-lan

提高MOEAs解集的分布性
——一种基于∞范数的逐步方法

曾映兰,郑金华,罗 彪   

  1. 湘潭大学 信息工程学院,湖南 湘潭 411105
  • 通讯作者: 曾映兰

Abstract: Diversity of solutions is one of the most important jobs of multi-objective optimization.Diversity includes the span and uniformity.In Multi-Objective Evolutionary Algorithms(MOEAs),population maintenance is used to realize the diversity.In this paper,a ∞-norm(infinite norm) based stepwise(INS) method is proposed to increase diversity of MOEAs.INS use ∞-norm as a measurement of diversity of individuals,and use stepwise method to wipe off individuals form population.Through experiments on 9 test problems,compared with two most popular MOEAs——NSGA-II and ε-MOEA,the experimental results demonstrate that INS can increase diversity of solutions obviously.

摘要: 解集的分布性是多目标优化中最重要的研究工作之一,解集的分布性主要体现在两个方面,一是解集的分布广度;二是解集的均匀性。在多目标进化算法(MOEAs)中,解集分布性的保持放在种群维护中实现,提出一种基于∞范数的逐步方法(INS)来提高MOEAs解集的分布性,INS用∞范数来衡量个体的分布性,用逐步的方法来裁剪个体。通过与目前最流行的两个MOEAs——NSGA-II和ε-MOEA,在9个测试函数上进行实验,结果表明INS能很好地提高解集的分布性。