BP神经网络的改进

BP神经网络易于陷入局部最小点以及收敛速度较慢，为了克服这些缺陷，本文对BP神经网络进行改进。通过对BP神经网络的样本进行采样分析，得到训练目标函数与输入向量之间的相关系数，依据此相关系数得到网络训练时的初始权重，再给待训练的BP神经网络进行初始权重的赋值，通过对初始权重的科学赋值从而达到避免网络在训练过程中陷入局部最小点与加快收敛速度的目的。本文通过实际验证，确实达到预期目的。

Improved BP Neural Network

This paper mainly aimed at avoiding local minimum of BP neural network and improving the convergence rate. Through the sample analysis of BP neural network, the correlation coefficient between the training objective function and the input vector was obtained, and the initial weights of the BP neural network were obtained by the correlation coefficient. And the experiments have achieved the expected goal.