构建BP网络结构方法及其在预测上应用  

Method of Constructing BP Network Structure and Its Application on Forecast

作　　者：孙红丽[1] 

机构地区：[1]商丘师范学院计算机科学系,河南商丘476000

出　　处：《河南科学》2009年第7期839-841,共3页Henan Science

基　　金：国家自然科学基金项目(50106011)

摘　　要：介绍了BP网络结构的重要性,总结了设计BP网时确定隐含层层数及其节点数的传统方法,并引入一种新方法,此方法将矩阵分解理论引入BP网络结构优化中,为BP网络构建指出了一个全新的思路及未来的发展趋势.将此理论应用与BP网络预测上,构建网络结构合理,预测结果精度高,实践应用效果良好,具有一定的推广意义.The importance of BP network structure is introduced in the paper, then several traditional methods and a new way about how to choose number of the hidden layers and its nodes when we design BP network are summarized, the new method involves the theory of matrix decomposition in optimizing the structure of back propagation network, and the author points out a overall new thought way, prospects the development trend in the future. At last, the result of applying the theory on the forecast is efficient, the structure of network is reasonable, the precision of forecast is high, the effect of practice is excellent, so the conclusion is worth spreading.

关 键 词：人工神经网络 BP网络 网络结构 奇异值分解 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]