输入学习样本，使用反向传播算法对网络的权值和偏差进行反复的调整训练，使输出的向量与期望向量尽可能地接近，当网络输出层的误差平方和小于指定的误差时训练完成，保存网络的权值和偏差。-Enter the study sample, using the back-propagation algorithm repeatedly adjust training of the network weights and deviation as close as possible, so that the output vector with expectation vector training is completed when the squared error of the network output layer is less than the specified error save the network weights and bias.