Journal of Systems Engineering and Electronics ›› 2010, Vol. 21 ›› Issue (5): 777-783.doi: 10.3969/j.issn.1004-4132.2010.05.010

• SYSTEMS ENGINEERING • Previous Articles     Next Articles

Multi-population and diffusion UMDA for dynamic multimodal problems

Yan Wu1,*, Yuping Wang2, Xiaoxiong Liu3, and Jimin Ye1   

  1. 1. School of Sciences, Xidian University, Xi’an 710071, P. R. China;
    2. School of Computer Science and Technology, Xidian University, Xi’an 710071, P. R. China;
    3. College of Automation, Northwestern Polytechnical University, Xi’an 710072, P. R. China
  • Online:2010-10-11 Published:2010-01-03

PDF (PC)

Abstract:

In dynamic environments, it is important to track changing optimal solutions over time. Univariate marginal distribution algorithm (UMDA) which is a class algorithm of estimation of distribution algorithms attracts more and more attention in recent years. In this paper a new multi-population and diffusion UMDA (MDUMDA) is proposed for dynamic multimodal problems. The multi-population approach is used to locate multiple local optima which are useful to find the global optimal solution quickly to dynamic multimodal problems. The diffusion model is used to increase the diversity in a guided fashion, which makes the neighbor individuals of previous optimal solutions move gradually from the previous optimal solutions and enlarge the search space. This approach uses both the information of current population and the part history information of the optimal solutions. Finally experimental studies on the moving peaks benchmark are carried out to evaluate the proposed algorithm and compare the performance of MDUMDA and multi-population quantum swarm optimization (MQSO) from the literature. The experimental results show that the MDUMDA is effective for the function with moving optimum and can adapt to the dynamic environments rapidly.   

京ICP备11035014号-5
Copyright © Editorial Office of Journal of Systems Engineering and Electronics, Beijing Institute of Aerospace Information
TEL: 86-10-68388406/68386014    E-mail：xtgcydzjs@126.com(系统工程与电子技术)/jseeoffice@126.com(JSEE)
Support by Beijing Magtech Co.ltd.