计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (19): 189-191.DOI: 10.3778/j.issn.1002-8331.2009.19.058

• 工程与应用 • 上一篇    下一篇

PID神经元网络之权值直接确定法研究

张雨浓1,杨逸文1,2   

  1. 1.中山大学 信息科学与技术学院,广州 510275
    2.中山大学 软件学院,广州 510275
  • 收稿日期:2009-01-09 修回日期:2009-02-19 出版日期:2009-07-01 发布日期:2009-07-01
  • 通讯作者: 张雨浓

On weights-direct-determination method of PID neurons’ network

ZHANG Yu-nong1,YANG Yi-wen1,2   

  1. 1.School of Information Science and Technology,Sun Yat-Sen University,Guangzhou 510275,China
    2.School of Software,Sun Yat-Sen University,Guangzhou 510275,China
  • Received:2009-01-09 Revised:2009-02-19 Online:2009-07-01 Published:2009-07-01
  • Contact: ZHANG Yu-nong

摘要: 网络权值如何确定,是人工神经网络研究中的一个重要课题。传统PID神经元网络在该问题的研究上,大多数采用误差回传(BP)的思想,通过迭代训练而估算出该网络的连接权值。针对PID神经元网络,对其进行简单巧妙的转化,可提出一种基于矩阵伪逆表述的直接计算权值的方法,从而避免了冗长的迭代训练过程。计算机仿真结果表明,该权值直接确定方法不仅有更快的学习/计算速度,而且能达到更高的计算精度。

关键词: PID神经元网络, 权值直接确定, 系统辨识, 矩阵伪逆

Abstract: Weights-determination is an important issue in the research of artificial neural networks.The traditional PID neurons’ network adopts mostly the standard error Back-Propagation(BP) algorithm which obtains the connecting weights via iterative training.In contrast,a pseudo-inverse based method is proposed,which could directly determine the neural-network weights just in one step without iterative BP training.Computer-simulation results show that this weights-direct-determination method could be more efficient and effective than standard BP-training algorithms.

Key words: PID neurons’ network, weights-direct-determination, system identification, pseudo-inverse