摘要
通过分析待学习参数的不同特性，提出一种快速收敛方法，使网络的收敛速度大大提高．通过对多种函数的实验，与不同的网络结果对比，表明此方法具有很强的普遍性．
A fast learning algorithm of neural network for approximation function was presented by analyzing different characters of parameters in network. This method can speed up the rate of network convergence. By approximating various functions, and comparing with other methods, this algorithm is shown to have satisfactory performances.
出处
《红外与毫米波学报》
SCIE
EI
CAS
CSCD
北大核心
1998年第4期303-307,共5页
Journal of Infrared and Millimeter Waves
基金
国家863基金
关键词
神经网络
函数逼近
子波分析
学习算法
neural network, approximating function, wavelet analysis, learning algorithm.
