摘要
对神经网络、小波网络和模糊小波网络函数逼近性进行对比,进而对采用正交化选择法和前向选择法净化小波时,小波网络和模糊小波网络对一维非线性函数逼近进行了分析。仿真结果证明模糊小波网络具有高精度的逼近能力和很强的泛化能力,该方法比小波网络和BP网络更优越,并且正交最小二乘法净化小波的性能指标优于前向选择法。
This paper compares the function approximation for BP neural network, wavelet neural network,fuzzy neural network.And for purifying the wavelets,the stepwise selection by orthogonalization and the residual based selection algorithm are compared when nonlinear functions are approximated using fuzzy neural network and neural network approach.The simulation results show that the approximating accuracy and generalization capability of FWN can be greatly improved.Moreover,the performance index for stepwise selection by orthogonalization is better than that of residual based selection algorithm.
出处
《西华大学学报(自然科学版)》
CAS
2008年第1期73-75,80,共4页
Journal of Xihua University:Natural Science Edition
基金
四川省教育厅重点项目(No.2005A117)
西华大学人才引进项目(R0620908)
关键词
函数逼近
神经网络
模糊小波神经网络
function approximation
neural network
fuzzy wavelet neural network