计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (20): 31-37.DOI: 10.3778/j.issn.1002-8331.1606-0137

• 理论与研发 • 上一篇    下一篇

动态邻居维度学习的多目标粒子群算法

肖闪丽,王宇嘉,聂善坤   

  1. 上海工程技术大学 电子电气工程学院,上海 201620
  • 出版日期:2017-10-15 发布日期:2017-10-31

Multi-objective particle swarm optimization based on dynamic neighborhood for dimensional learning

XIAO Shanli, WANG Yujia, NIE Shankun   

  1. School of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201620, China
  • Online:2017-10-15 Published:2017-10-31

摘要: 针对多目标粒子群算法多样性较差,种群选择压力随着变量维度增加的问题,提出了基于动态邻居维度学习的多目标粒子群算法(DNDL-MOPSO)。该算法首先构建最优维度个体,然后在“个体认知”和“社会认知”的基础上,对粒子速度更新公式进行改进,采用每一维上学习对象不固定的交流方式,最后利用随机向导学习策略,增加种群多样性。实验结果表明该方法能够提高算法的全局收敛性,增加种群的多样性,缓解选择压力,有效解决多峰多目标优化问题。

关键词: 粒子群算法, 多目标优化, 动态邻居, 最优维度粒子, 随机向导学习

Abstract: Focus on the poor behavior of the diversity for multi-objective particle swarm optimization and the selection pressure of population increasing with the variable dimension, a Multi-Objective Particle Swarm Optimization based on Dynamic Neighborhood of Dimensional Learning (DNDL-MOPSO) is proposed. Firstly, an optimum dimensional individual is established. Then based on the individual and social knowledge, the proposed algorithm improves the formula of the velocity updating and uses a strategy that each dimensional learning object is not fixed. Finally, the random guide learning strategy is used to alleviate the selection pressure. The experimental results indicate that the new algorithm can improve the global convergence and increase the diversity of population. It is effective to solve the benchmark multimodal optimization problems.

Key words: Particle Swarm Optimization(PSO), multi-objective optimization, dynamic neighbor, optimum dimensional individual, random guide learning