摘要： 针对目前常用的多层前馈神经元网络的BP学习算法的收敛速度慢、易出现病态情况、且容易陷入局部极小点的缺点,提出了一种新的多层前馈神经元网络训练算法。该算法具有指数收敛速度,能克服病态,算法简单,易于应用,并且给出了该算法收敛性的证明。实例分析验证了算法的有效性。

关键词: 多层前馈神经元网络, 指数收敛速度, 训练算法

Abstract: This paper presents a new multilayer feedforward neural networks learning algorithm. It is of exponential convergence speed and also applicable for the illconditioned stuations.This algorithm is simple, easy to use, and its convergence is discussed, too.Example result indicate effectiveness of thealgorithm.

中图分类号: 

• TP183