一种快速BP网络训练算法及应用

针对基本BP算法在接近最优点时收敛速度变慢的缺点，提出一种快速BP网络训练算法。该算法将网络训练过程分为两个阶段，首先采用变学习率BP算法使其接近最优点，再改用PRP共轭梯度法以加快收敛速度，其中学习率通过不精确一维搜索方法寻找得到。并采用Widrow—Nguyen初始化规则选取初始权值。通过两个仿真试验表明该算法是一种高效、快速的BP网络训练方法。同时，该算法具有简单、易于实现的优点，具有较为广阔的应用前景。

A Fast Learning Algorithm of BP Network and Its Application

To overcome the defect of basic BP algorithm, this paper proposed a new faster learning algorithm, which divides the training process into two phases. First it adapts BP algorithm to approach the best point , then changes gradient descent algorithm to quicken the speed. Simulation results indicate that the proposed method is an efficient and fast algorithm.