本文已被：浏览 1098次 下载 455次
投稿时间：2008-02-08
投稿时间：2008-02-08
中文摘要: 神经网络模型是一种非常有效的数据处理工具,但是存在结构确定困难的缺点.针对神经网络算法的这种缺点,提出了变结构神经网络模型.此模型增加了神经网络隐节点的决策变量,并对此决策变量进行松弛.在采用BP梯度算法确定神经网络结构的同时,确定网络参数.由于电缆的状态监测是时序数据,将此模型应用于电缆的状态监测过程中,能体现出较好的适应性.
Abstract:Neural network model is an effective data processing method, but it has a shortcoming of difficult structure confirmation. This paper aims at this primary disadvantage of the neural network calculation, giving a variable structure neural network model, which model increases decisionmaking variable in allusion to the hidden nodes of neutral network, fixing network parameter while using BP gradient algorithm to fix the structure of neural network by means of slacking this decisionmaking variable. In applying this model to cable state detection which is time series data, in processing continuous changing data, this model has good adaptability.
keywords: neural network BP calculation state detection
文章编号：20080116 中图分类号： 文献标志码：
基金项目:上海市科学技术委员会西部项目资助计划
引用文本：
方先存,王承民,张铁岩.变结构神经网络模型的BP算法及其应用研究[J].上海电力大学学报,2008,24(1):57-60,64.
FANG Xian-cun,WANG Cheng-min,ZHANG Tie-yan.BP Calculation of Variable Structure Neural Network Model and Its Application[J].Journal of Shanghai University of Electric Power,2008,24(1):57-60,64.
方先存,王承民,张铁岩.变结构神经网络模型的BP算法及其应用研究[J].上海电力大学学报,2008,24(1):57-60,64.
FANG Xian-cun,WANG Cheng-min,ZHANG Tie-yan.BP Calculation of Variable Structure Neural Network Model and Its Application[J].Journal of Shanghai University of Electric Power,2008,24(1):57-60,64.