摘要
前馈神经网已经被大量用于非线性信号处理 .经典反向传播算法是一种标准的前馈网络学习算法 ,但是 ,对许多应用 ,反向传播算法的收敛速度却很慢 .本文根据对网络的非线性单元进行线性化而提出一种新的算法 ,该算法在非线性信号处理中在精度和收敛速度方面都优于传统的反向传播算法 .
The feedforward neural networks has been greatly used in the nonlinear signals processing. The classical BP algorithm is a standard learning algorithm for the feedforward neural netwrks,but in many applications, the BP algorithm has a very slow convergence speed.In this paper,a new algorithm on the linearization of nonlinear neural in feedforward neural networks is proposed,and its precision and convergence speed is better than the classical BP algorithm's.
出处
《信息与控制》
CSCD
北大核心
2000年第1期34-39,共6页
Information and Control
关键词
快速算法
前馈神经网
线性化
信号处理
fast algorithm, feedforward neural networks, linearization
