计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (34): 39-43.

• 研究、探讨 • 上一篇    下一篇

基于ε-支配的自适应多目标进化算法

梁 浩,林 丹,马 楠   

  1. 天津大学 理学院 数学系,天津 300072
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-12-01 发布日期:2011-12-01

ε-dominance based adaptive multi-objective evolutionary algorithm

LIANG Hao,LIN Dan,MA Nan   

  1. Department of Mathematics,School of Science,Tianjin University,Tianjin 300072,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-12-01 Published:2011-12-01

摘要: 提出一种新的基于ε-支配关系的自适应多目标进化算法(AEMOEA)。在每次的进化中保留端点,并从端点集中选取一个作为父本,参加进化,弥补了ε-MOEA算法中端点易被丢掉的缺陷;在进化过程中根据存档动态地调整ε的取值,使解的分布更加均匀;当存档中个体过多时,运用ε-支配关系进行剪切,使其个体数处在合理水平。通过5个常用双目标测试函数的计算,验证了该算法在求解质量上优于ε-MOEA、NAGA-II以及SPEA-2等主流多目标算法。

关键词: 多目标优化, 多目标进化算法, ε-支配, ε-自适应调整

Abstract: A novel multi-objective evolutionary algorithm,called ε-dominance based adaptive multi-objective evolutionary algorithm(AEMOEA),is proposed.As the improvement to ε-MOEA,the boundary points easily discarded before are reserved in AEMOEA and one of them is chosen as parent to take part in each evolution.In addition,the εvalue is dynamically modified with the archive in every generation to find a well-distributed set of solutions.Finally,when the archive is over-sized,ε-dominance is used to reduce it to a proper number.The proposed AEMOEA algorithm is tested on five bi-objective benchmark test functions,and the experimental results demonstrate that AEMOEA outperforms other MOEAs such as ε-MOEA,NAGA-II and SPEA-2.

Key words: multi-objective optimization, multi-objective evolutionary algorithm, ε-dominance, ε-adaptive