一种改进的BP算法

An Improvement in BP Algorithm

讨论了网络学习过程中的假饱和现象，并给出了克服方法，同时，也讨论了学习样本输入编排机理，给出了一种避免网络学习出现的局部极小的算法。

This paper discusses fault saturation in studying process of neural network,and presents a method of avoiding fault saturation.Meanwhile,the paper also discusses the arranging mechanism of input sample of neural network,and presents a algorithm to avoid local minimum in studying process of neural network.Practical example indicates that the method work well.