Computer Engineering and Applications ›› 2008, Vol. 44 ›› Issue (23): 42-45.DOI: 10.3778/j.issn.1002-8331.2008.23.0013

• 理论研究 • Previous Articles     Next Articles

Density clustering based niching Differential Evolution

ZHANG Hang1,WANG Wei1,ZHENG Ling2,LI Dan-dan3,XIONG Fu-qiang1   

  1. 1.School of Information Science and Engineering,Central South University,Changsha 410083,China
    2.Department of Accounting,Hunan Business College,Changsha 410008,China
    3.College of Finance,Hunan University,Changsha 410079,China
  • Received:2008-03-31 Revised:2008-05-28 Online:2008-08-11 Published:2008-08-11
  • Contact: ZHANG Hang

一种基于密度聚类的小生境差分进化算法

张 航1,王 伟1,郑 玲2,李丹丹3,熊富强1   

  1. 1.中南大学 信息科学与工程学院,长沙 410083
    2.湖南商学院 会计系,长沙 410008
    3.湖南大学 金融学院,长沙 410079
  • 通讯作者: 张 航

Abstract: Considering the premature convergence problem in the conventional differential evolution algorithm,a density clustering based niching Differential Evolution algorithm is proposed in this paper.Based on the strong global searching ability and good robustness of DE/rand/2/bin mutation scheme and the good performance of local searching ability and fast convergence speed of DE/best/2/bin mutation scheme,the algorithm initializes a global population without sub-populations at first,then iteratively searches the global population employing DE/rand/2/bin scheme and makes clustering to individuals,a niche sub-population forms when the individual amounts in cluster reach the smallest size specified,and then iteratively searches sub-populations employing the improved DE/best/2/bin scheme and makes clustering again,so as to improve the population diversity in the process of evolution,and strengthen the ability of avoiding local optimization.Simulation result shows that the algorithm can greatly improve the convergence speed and the global searching ability,and efficiently avoid premature convergence.

Key words: premature convergence, density clustering, niching, Differential Evolution(DE), population diversity

摘要: 针对基本差分进化算法早熟收敛的缺陷,提出了一种基于密度聚类的小生境差分进化算法。该算法基于DE/rand/2/bin变异方式全局搜索能力强、鲁棒性好和DE/best/2/bin变异方式局部搜索能力强、收敛速度快的特点,首先初始化一个没有子种群的全局种群,再在全局种群中采用DE/rand/2/bin进行迭代搜索,并对其中的个体进行聚类,当聚类簇中的个体数目达到规定的最小规模时形成一个小生境子种群,然后在各子种群中采用改进的DE/best/2/bin进行迭代搜索并重新进行聚类,从而提高进化过程中种群的多样性,增强算法跳出局部最优的能力。仿真实验表明,该方法能显著提高算法的收敛速度和全局搜索能力,有效避免早熟收敛。

关键词: 早熟收敛, 密度聚类, 小生境, 差分进化, 种群多样性