神经网络BP学习算法的改进

AN MODIFIED LEARNING ALGORITHM OF BP NEURAL NET

神经网络的BP（误差反向传播）学习不法是针对多层前债结构的神经网络进行研究而·提出的一种有效的学习算法。它因具有良好的自适应、自学习及容错、抗干扰能力而受到广泛应用。木文对BP算法收敛速度慢的原因进行了分析，讨论了一些提高BP算法收敛速度的改进措施，并第一次提出了一种新的改进BP学习算法一多学习年学习算法。

The back propagation (BP) learning algorithm of Artificial Neural Net was briefly reviewed. Thereason that the original BP algorithm has lower convergence speed was analysed and some modifications of BP al-gorithm were discussed. A new modification of BP learning algorithm was proposed and tested by several traininggroup compared to the original BP algorithm and its other modifications. The experimental results showed thatthe proposed algorithm revealed higher training speed and accuracy.