基于QR-ROLS算法的小波神经网络应用研究  

APPLICATION OF WAVELET NEURAL NETWORK BASED ON QR-ROLS ALGORITHM

作　　者：黄宜军[1,2] 李艳灵[1] 

机构地区：[1]信阳师范学院计算机系,河南信阳464000 [2]空军第一航空学院,河南信阳464000

出　　处：《计算机应用与软件》2008年第11期245-246,258,共3页Computer Applications and Software

摘　　要：针对小波神经网络的结构设计和权值选取,提出一种基于QR分解的递推正交最小二乘算法以节省计算资源和计算时间,仿真结果表明基于QR-ROLs(QR Decomposition-Recursive Orthogonal Least Squares)算法的小波神经网络比依据时频特性设计的小波神经网络结构更优,运行效率更高。Considering the structural design and weight selection of wavelet neural network, a recursive orthogonal least squares algorithm based on QR decomposition is presented. The computing time and resources are saved by this algorithm. Simulation results show that the wavelet network based on QR-ROLS algorithm has better structure and works more efficiently than the wavelet network designed by time-frequency property of signal.

关 键 词：小波分析 神经网络 QR分解 RLS算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP393[自动化与计算机技术—控制科学与工程]