一种改进的BP算法

An Improved BP Algorithm and its Application

为了克服BP算法易陷入局部极小、收敛速度慢等缺点,笔者利用非线性最小二乘法对其进行了改进.结果表明,采用改进后的BP算法来训练神经网络,能在一定程度上提高神经网络的收敛速度,具有学习速度快、识别能力强等优点.

BP(back propagation) algorithm encounters local minimum, slow convergence speed and convergence instability. The shortcomings can be overcome by application of the nonlinear least square method in this paper. Application result shows that rohen neural network is trained by the modified BP algorithm after improvement, the convergency speed can be increased to a certain extent,and it has many advantages such as fast speed of learning, better capability of recongnition.