计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (11): 162-164.DOI: 10.3778/j.issn.1002-8331.2009.11.049

• 数据库、信号与信息处理 • 上一篇    下一篇

遗传算法优化神经网络权值盲均衡算法的研究

张立毅1,2,刘 婷1,孙云山1,李 锵2   

  1. 1.天津商业大学 信息工程学院,天津 300134
    2.天津大学 电子信息工程学院,天津 300072
  • 收稿日期:2008-03-03 修回日期:2008-05-07 出版日期:2009-04-11 发布日期:2009-04-11
  • 通讯作者: 张立毅

Research of genetic algorithm optimization neural network weights blind equalization algorithm based on real number coding

ZHANG Li-yi1,2,LIU Ting1,SUN Yun-shan1,LI Qiang2   

  1. 1.College of Information Engineering,Tianjin University of Commerce,Tianjin 300134,China
    2.School of Electric Information Engineering,Tianjin University,Tianjin 300072,China
  • Received:2008-03-03 Revised:2008-05-07 Online:2009-04-11 Published:2009-04-11
  • Contact: ZHANG Li-yi

摘要: 将遗传算法与神经网络盲均衡算法相结合,提出了两段式优化神经网络权值的方案。首先利用遗传算法全局搜索能力强的特点优化初始权值,然后发挥BP算法局部搜索速度快的特点得到最佳权值。经计算机仿真表明,该算法与传统BP神经网络盲均衡算法相比,收敛速度加快,稳态剩余误差减小,误码率降低。

关键词: 盲均衡算法, 神经网络, 遗传算法, 初始权值

Abstract: The project of two-stage optimization neural network weights is proposed by combining genetic algorithm with neural network blind equalization algorithm.At first,the initialization weight is optimized using the characteristic of genetic algorithm,which is strong global search capability.And then,optimal weight is gained in virtue of the merit of BP algorithm,which is fast local search speed.Computer simulations show that,compared with traditional blind equalization algorithm based on BP neural network,the convergence speed of the proposed algorithm is quickened,state residual error is decreased and BER is reduced.

Key words: blind equalization algorithm, neural network, genetic algorithm, initial weight