BP神经网络基本原理概述：这种网络模型利用误差反向传播训练算法模型，能够很好地解决多层网络中隐含层神经元连接权值系数的学习问题，它的特点是信号前向传播、误差反向传播，简称BP（Back　Propagation）神经网络。BP学习算法的基本原理是梯度最快下降法，即通过调整权值使网络总误差最小，在信号前向传播阶段，输入信号经输入层处理再经隐含层处理最后传向输出层处理；在误差反向传播阶段，将输出层输出的信号值与期望输出信号值比较得到误差，若误差较大则把误差信号传回隐含层直至输入层，在各层神经元中使用误差信号修改权值系数，之后进入下一轮迭代，如此循环直至误差最小，实际输出信号值接近期望输出信号值。下图为三层BP神经网络模型：(In pattern recognition, there is a highly practical classification method, which is artificial neural network. It has been successfully applied to intelligent robot, automatic control, speech recognition, prediction estimation, biology, medicine, economy and other fields. It has solved many practical problems which are difficult to solve by many other classification methods. This is due to the many models of neural network, and the corresponding neural network model can be used for different problems. The BP neural network is used to solve the problem of handwritten digital character recognition.)