作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2007, Vol. 33 ›› Issue (05): 164-165. doi: 10.3969/j.issn.1000-3428.2007.05.058

• 人工智能及识别技术 • 上一篇    下一篇

一种免疫克隆选择算法在BP网络学习中的应用

姜新农,王文香   

  1. (中国科技大学计算机科学技术系,合肥 230027)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-03-05 发布日期:2007-03-05

Application of Immune Clonal Selection Algorithm in BP Network Learning

JIANG Xinnong, WANG Wenxiang   

  1. (Department of Computer Science and Technology, University of Science and Technology of China, Hefei 230027)
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-03-05 Published:2007-03-05

摘要: 在分析BP网络学习存在的问题后,采用了一种免疫克隆选择算法对BP网络的权值进行优化学习,并提出了一种新的变异方法,该变异方法可以根据亲和力的大小自适应调整抗体变异的幅度,与传统的高斯变异相比,不但简化了抗体的编码,还很好地体现了克隆选择算法抗体变异的特点,提高了算法的搜索能力和收敛性能。仿真实验表明,基于这种变异方法的免疫克隆选择算法可以很好地提高BP网络的学习速度,有效地避免算法过早收敛的问题。

关键词: BP网络, 免疫克隆选择算法, 变异, 运动学逆问题

Abstract: After analyzing the shortcomings of learning BP network simply, the paper applies an immune clonal selection algorithm(ICSA)to optimizing the weights of BP network, and puts forward a new method of mutation. The method of mutation can adjust the extent of mutation adaptively by the affinity of antibody, in comparison with Gauss mutation, it can not only simplify the coding of antibody, but also incarnate the characteristic of mutation of ICSA and improve the capability of search and convergence of algorithm. The simulation results show that ICSA based on this method of mutation can improve the rapidity of learning BP network well and avoid prematurity effectively.

Key words: BP network, Immune clonal selection algorithm, Mutation, Inverse kinematics