摘要
基于函数逼近,对前向神经元网络中非线性函数与子波变换中母函数的关系进行了分析,推出了前向神经元网络的子波表示.指出两种方法应用于L2(R)空间函数逼近时是一致的.最后给出了两个前向神经元网络的子波表示,函数逼近实例验证了本文结论的正确性.
Based on the approximation theory of functions, this paper studies the relationship between the nonlinear function of feedforward neural network and the mother function of wavelet transform. It is pointed out that the two kinds of function approximation approaches in L2(R) spaces are equivalent, and the wavelet representation of feedforward neural network is then given. In the last part of the paper,examples of wavelet representations of feedforward neural network and function approximation are shown.
出处
《北京理工大学学报》
EI
CAS
CSCD
1995年第3期305-310,共6页
Transactions of Beijing Institute of Technology
基金
高等学校博士学科点专项科研基金
电子科学研究院基金
关键词
神经网络
前向神经元网络
子波变换
neural networks
frames/approximation of function
feedforward neural network
wavelet transform