English

应用科学学报 ›› 2011, Vol. 29 ›› Issue (3): 308-315.

• 信号与信息处理 • 上一篇    下一篇

基于当前最优解的反向差分进化算法求解函数优化问题

徐庆征1;2， 王磊1， 何宝民1， 王娜2   

  1. 1. 西安理工大学计算机科学与工程学院，西安710048
    2. 西安通信学院军事电子工程系，西安710106
  • 收稿日期:2010-12-07 修回日期:2011-04-10 出版日期:2011-05-26 发布日期:2011-05-26
  • 作者简介:徐庆征，博士生，讲师，研究方向：进化计算，E-mail: xuqingzheng@hotmail.com；王磊，教授，博导，研究方向：自然计算、智能信息处理，E-mail: leiwang@xaut.edu.cn
  • 基金资助:

    国家自然科学基金(No.60802056, No.61073091)；陕西省自然科学基金(No.2010JM8028)；西安理工大学优秀博士学位论文研
    究基金(No.105-211010)资助

Opposition-Based Differential Evolution Using the Current Optimum for Function Optimization

XU Qing-zheng1;2, WANG Lei1, HE Bao-min1, WANG Na2   

  1. 1. School of Computer Science and Engineering, Xi’an University of Technology, Xi’an 710048, China
    2. Department of Military Electronic Engineering, Xi’an Communication Institute, Xi’an 710106, China
  • Received:2010-12-07 Revised:2011-04-10 Online:2011-05-26 Published:2011-05-26

PDF

2071

被引次数

13

摘要：

当最优解偏离目标函数定义域的几何中心时，反向个体容易远离全局最优解，基于反向差分进化算法的性能会大幅降低. 该文引入基于当前最优解的反向学习策略，并与差分进化算法相结合，求解函数优化问题. 当前代的最优解作为候选解和相应反向个体之间的对称点，能保证反向种群的利用率始终维持在较高水平. 实验结果表明，该算法可行而高效，且算法性能的提升完全是反向个体的贡献. 此外，提出一种增强的基于反向差分进化算法，展示出此类优化方法的最优效果.

关键词: 差分进化, 基于反向学习, 当前最优解, 函数优化

Abstract:

When the global optimum is not located at the geometric center of the domain, the opposite numbers may lapse from the global optimum, leading to poor performance of opposition-based differential evolution. A novel opposition-based learning strategy using the current optimum is introduced, and it is combined with differential evolution for function optimization. The optimum in the current generation is served as a symmetry point between an estimate and the corresponding opposite estimate, resulting in a high rate of opposite population usage. Experiments results clearly show that the proposed algorithm can significantly improve the performance due to the opposite numbers. Additionally, an enhanced version of opposition-based
differential evolution is proposed to reveal ideal and perfect results using opposition-based learning.

Key words: differential evolution, opposition-based learning, current optimum, function optimization

中图分类号: