一种提高BP神经网络泛化能力的改进算法  

A Modified Algorithm to Improve Generalization Ability of BP Neural Network

作　　者：夏玫[1] 陈立潮[1] 王新波[1] 

机构地区：[1]太原科技大学计算机科学与技术学院,山西太原030024

出　　处：《计算机技术与发展》2009年第9期62-64,68,共4页Computer Technology and Development

基　　金：山西省科技攻关项目资助(20080321012-01)

摘　　要：针对BP神经网络中样本数据复杂和容易陷入到局部极值的不足,通过利用主成分分析法对样本数据进行简化,并采用遗传算法优化神经网络的初始权值,提出了一种改进BP神经网络泛化能力的算法。该算法兼顾了主成分分析方法剔除冗余的特性和遗传算法能够全局优化的特点,既具备了较快的收敛速度又不易陷入到局部极值中,具有较强的泛化能力。实验结果表明,改进后的算法在数据降维、冗余信息的剔除、预测能力等方面都得到了一定程度上的改善。To solve the complexity of the sample data of BP neural network and easily getting into a local value, an improved algorithm is proposed in this paper. This algorithm uses the principle components analysis to predigest the sample and the genetic algorithms to find the best point, and it combines the eharaeter of PCA to reduce redundancy with that of GA to optimize the whole, so it can not only pick up convergence and can avoid to get into a local value. With the result of the experiment of the improved algorithm of BP neural network, it is valid to reduce the dimensions of data, delete redundancies, and increase forecast ability.

关 键 词：BP神经网络 泛化能力 主成分分析 遗传算法 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]