摘要
在小波分析的基础上 ,结合神经网络构造了一种新的小波级数形式的小波网络 ,并利用非线性规划中的变尺度法对网络权值进行训练。对一类非线性函数和混沌时间序列进行了仿真研究 ,并与文献 [7]的小波网络进行了仿真比较。结果表明 ,这种新的小波网络具有良好的函数逼近能力 ,收敛速度快且逼近精度高。
In this paper, based on the wavelet analysis and combined with neural network, a novel wavelet series typed wavelet network is constructed. The variable-schedule method is introduced to adjust the weights. Simulations on a type of nonlinear function and a chaotic time series are studied. The comparison with the reference [7] is also given. Simulation results indicate that the proposed wavelet network in the paper has a good approximation capability, and has the advantages of rapid convergence rate and high approximation accuracy.
出处
《系统工程与电子技术》
EI
CSCD
2000年第5期43-45,86,共4页
Systems Engineering and Electronics
基金
高校博士点基金资助课题!(BJ970 3 )
