摘要
在研究MADALINES人工神经网络中心差梯度学习算法(CDG)的基础上,提出了一种加速中心差梯度学习算法(SCDG),给出了该算法收敛性的证明,在计算机上模拟实现了SCDG算法且分析了实验结果,理论分析与实验均表明,SCDG算法较之传统的CDG算法,收敛速度提高了二个数量级以上.
Based on the Central-Difference-Gradient Learning Algorithm (CDG) used in MADA-LINES artificial neural networks, a speed Central-Difference-Gradient Learning Algorithm (SCDG) is proposed and the convergency demonstrated. The simulation of this algorithm is performed on a type T&W386 computer and the experimental results are discussed. It has been proved that the convergence rate of SCDG algorithm can be improved by more than two orders of magnitude as compared with that of the conventional one.
出处
《华中理工大学学报》
CSCD
北大核心
1993年第6期6-10,共5页
Journal of Huazhong University of Science and Technology
关键词
学习算法
语音处理
人工神经网络
ANN
learning algorithm
pattern recognition
speech processing