首页 | 本学科首页   官方微博 | 高级检索  
     检索      

小波神经网络的参数初始化研究
引用本文:赵学智,邹春华,陈统坚,叶邦彦,彭永红.小波神经网络的参数初始化研究[J].华南理工大学学报(自然科学版),2003,31(2):77-79,84.
作者姓名:赵学智  邹春华  陈统坚  叶邦彦  彭永红
作者单位:1. 华南理工大学,机械工程学院,广东,广州,510640
2. 广州(从化)亨龙机电制造实业有限公司,广东,从化,510990
基金项目:国家自然科学基金资助项目(59905008),广东省自然科学基金资助项目(980396),华南理工大学自然科学基金资助项目(E5305292)
摘    要:随机产生的初始参数往往使小波神经网络的学习次数大幅度地增加,甚至不收剑,为了加快网络的学习速度,本研究提出了一种将小波网络的初始参数设置和小波类型,小波时频参数和学习样本等联系起来的小波神经网络的初始参数设置方法,学习实例结果表明,按照这一方法不但可以获得高几率的优秀初始参数,而且能大大加快小波网络的后续学习速度。

关 键 词:小波神经网络  参数初始化  小波时频参数  学习样本  小波理论  人工神经网络
文章编号:1000-565X(2003)02-0077-04

A Research on the Initialization of Parameters of Wavelet Neural Networks
Zhao Xue-zhi Zou Chun-hua Chen Tong-jian Ye Bang-yan Peng Yong-hong.A Research on the Initialization of Parameters of Wavelet Neural Networks[J].Journal of South China University of Technology(Natural Science Edition),2003,31(2):77-79,84.
Authors:Zhao Xue-zhi Zou Chun-hua Chen Tong-jian Ye Bang-yan Peng Yong-hong
Abstract:Parameters obtained at random tend to increase the training times of wavelet neural networks and even make the training course unconvergent. In order to accelerate the training speed of the wavelet neural networks,a method of how to set the initial parameters of the wavelet neural networks is proposed in this paper. This method integrates the setting of initial parameters with the wavelet type,time-frequency parameters of the wavelet and the training samples. The training example shows that super initial parameters can be obtained with high probability by this method and, as a result, the network' s training speed can be accelerated very quickly.
Keywords:wavelet neural network  initialization of parameter  wavelet time-frequency parameter  training sample
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号