电子学报 ›› 2016, Vol. 44 ›› Issue (5): 1071-1077.DOI: 10.3969/j.issn.0372-2112.2016.05.009

• 学术论文 • 上一篇    下一篇

一种改进的基于密度的多目标进化算法

王鹏, 张长胜, 张斌, 刘婷婷   

  1. 东北大学信息科学与工程学院, 辽宁沈阳 110819
  • 收稿日期:2014-10-28 修回日期:2015-03-26 出版日期:2016-05-25 发布日期:2016-05-25
  • 通讯作者: 张斌
  • 作者简介:王鹏 男,1987年生于山东烟台.东北大学计算机应用技术专业博士研究生.研究方向为服务计算、人工智能算法;张长胜 男,1980年生于吉林长春.东北大学信息科学与工程学院副教授、硕士生导师.主要研究方向为智能信息处理.
  • 基金资助:

    宁夏回族自治区自然科学基金(No.NZ13265);中央高校东北大学基本科研专项基金(No.N120804001,No.N120204003)

An Improved Density-Driven Multi-objective Evolutionary Algotithm

WANG Peng, ZHANG Chang-sheng, ZHANG Bin, LIU Ting-ting   

  1. College of Information Science & Engineering, Northeastern University, Shenyang, Liaoning 110819, China
  • Received:2014-10-28 Revised:2015-03-26 Online:2016-05-25 Published:2016-05-25

摘要:

多目标密度驱动进化算法(MODdEA)利用非支配等级信息和分区密度信息求解多目标优化问题,该算法在与其他多目标进化算法的比较中有着出色的表现.在其基础上本文提出了一种改进的多目标进化算法MODdEA+,首先在该算法中基于搜索空间的分区机制提出了克隆操作,该操作不但能在进化前期增强算法的全局搜索能力,还能在进化后期提高算法的局部精化能力;其次引入一种基于Pareto信息表中个体支配及被支配信息的评价策略以使对信息表个体的排序结果更加精确;最后对变异操作进行了改进以降低出现不必要越界情况的概率.为验证改进算法的有效性,在对其进行分析的基础上针对多个测试问题将其与原算法进行了实验比较,结果表明改进算法的求解质量明显优于原算法.

关键词: 进化算法, 密度驱动, 克隆操作, 粗适应度值, 变异操作

Abstract:

Multi-objective evolutionary algorithm that diversifies population by its density (MODdEA) solve multi-objective optimization problem according to the non-dominated sorting information and spatial density information, the algorithm has a good performance in the comparison with other multi-objective evolutionary algorithm.In this paper, we propose an improved multi-objective evolutionary algorithm MODdEA + based on MODdEA.Firstly, we propose a operator named clone operator based on the partition mechanism in search space, this operator could not only improve the global search capabilities in the early stage of evolution, but also enhance the local refinement capabilities in the late stage of evolution;secondly, we introduce a evaluation strategy which evaluate the individuals in Pareto information list based on the dominate and dominated information, this strategy provide a more accurate sorting result;finally, we improve the mutation operator in order to reduce the probability of overstep of the boundary.To demonstrate the effectiveness of the improved algorithm, we compare it with MODdEA on multiple testing problems, the experimental results show that the improved algorithm's solving quality is much better than the original algorithm's.

Key words: evolutionary algorithm, density-driven, clone operator, raw fitness, mutation operator

中图分类号: