一种基于BP神经网络的数字识别新方法

针对标准BP神经网络收敛速度慢、易陷入局部极小点的缺点,提出了一种新的BP神经网络改进算法。该算法通过变步长法和牛顿法来改进BP算法,加快了网络的收敛速度,且收敛速度快于其他的改进算法。在此基础上将BP神经网络应用于数字识别中,为其网络建立识别模型。利用仿真实验观察BP网络的泛化能力以及识别准确性,比较BP算法及其改进方案,提出改进方案中分别需要注意的地方。

A new method of numerical recognition based on improved BP neural network

In this paper, a new improved BP algorithm is presented to solve the problem that BP neural network can easily fall into slow convergence and minimum. This algorithm integrates the variable step method with the Newton method. The algorithm can speed up the convergence rate of BP neural network, and the convergence rate is faster than other algorithm. The BP neural network is used in the digital recognition and the identification model for the network is established. Using simulation experiments to observe generalization ability and identification accuracy of BP neural network, compare BP neural network with its improvement scheme, and put forward the improvement plan in note respectively place.