电子学报 ›› 2016, Vol. 44 ›› Issue (7): 1643-1648.DOI: 10.3969/j.issn.0372-2112.2016.07.018

• 学术论文 • 上一篇    下一篇

并行协作骨干粒子群优化算法

申元霞1, 曾传华2, 王喜凤1, 汪小燕1   

  1. 1. 安徽工业大学计算机科学与技术学院, 安徽马鞍山 243032;
    2. 安徽工业大学数理科学与工程学院, 安徽马鞍山 243032
  • 收稿日期:2015-04-07 修回日期:2015-06-21 出版日期:2016-07-25
    • 通讯作者:
    • 申元霞
    • 作者简介:
    • 曾传华 男,1975年生于重庆,讲师,研究方向:概率与统计理论、偏微分方程数值求解.E-mail:stonezch@163.com;王喜凤 女,1980年生于山东成武,讲师,研究方向:信息安全、智能信息处理.E-mail:wxf80106@163.com;汪小燕 女,1974年生于安徽桐城,硕士,副教授,研究方向:数据挖掘、粗糙集理论.E-mail:wxyzjx@126.com
    • 基金资助:
    • 国家自然科学基金 (No.61300059,No.61472056); 安徽高校省级自然科学基金 (No.KJ2012Z031,No.KJ2012Z024)

A Parallel-Cooperative Bare-Bone Particle Swarm Optimization Algorithm

SHEN Yuan-xia1, ZENG Chuan-hua2, WANG Xi-feng1, WANG Xiao-yan1   

  1. 1. School of Computer Science and Technology, Anhui University of Technology, Maanshan, Anhui 243002, China;
    2. School of Mathematics & Physics, Anhui University of Technology, Maanshan, Anhui 243002, China
  • Received:2015-04-07 Revised:2015-06-21 Online:2016-07-25 Published:2016-07-25
    • Supported by:
    • National Natural Science Foundation of China (No.61300059, No.61472056); Provincial Natural Science Fund of Colleges and Universities of Anhui Province (No.KJ2012Z031, No.KJ2012Z024)

摘要:

为解决骨干粒子群优化(Bare-Bone Particle Swarm Optimization,BBPSO)的早期收敛问题,本文通过粒子的运动行为分析了导致BBPSO早期收敛的因素,并提出并行协作BBPSO,该算法采用并行的主群和从群之间的协作学习来平衡勘探和开采能力.为了增强主群的勘探能力,提出动态学习榜样策略以保持群体多样性;同时提出随机反向学习机制以实现从群的从全局到局部的自适应搜索功能.在14个不同特征的测试函数上将本文算法与6种知名的BBPSO算法进行对比,仿真结果和统计分析表明本文算法在收敛速度和精度上都有显著提高.

关键词: 骨干粒子群优化, 协作学习, 反向学习, 多样性

Abstract:

To deal with the premature convergence of the bare-bone particle swarm optimization (BBPSO) algorithm,we make the analysis of the motion behavior of the particles and point out the reasons leading to the premature convergence.According to the analysis results,a parallel-cooperative BBPSO (PCBBPSO) algorithm is proposed in which the parallel-cooperative learning of a master swarm and a slave swarm balances between exploration and exploitation abilities.In order to improve the exploration ability of the master swarm,a dynamic learning exemplar strategy is presented to preserve the swarm diversity.Meanwhile,a stochastic opposition-based learning mechanism is developed to achieve the abilities of the slave swarm from the global search to the local search.The proposed algorithm was evaluated on 14 benchmark functions with different characteristics.The experimental results and statistic analysis show that the proposed method significantly outperforms six state-of-the-art BBPSO variants in terms of convergence speed and solution accuracy.

Key words: BBPSO, cooperative learning, opposition-based learning, diversity

中图分类号: