QR-ROLS算法在小波网络构造中的仿真研究  

Research on Forming Wavelet Network in QR-ROLS Algorithm

作　　者：黄宜军[1] 刘宏兵[1] 熊炎[1] 

机构地区：[1]信阳师范学院计算机系

出　　处：《系统仿真学报》2007年第9期2101-2103,共3页Journal of System Simulation

摘　　要：针对小波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提高构造小波网络效率。仿真结果表明,QR-ROLS算法可有效地设计小波神经网络结构并同时对其进行优化。Wavelet network is the application of wavelet analysis in neural network fields. It always suffered from the curse of dimensionality. The number of weights needed increases exponentially with the number of inputs and outputs. Many methods have been proposed to solve this problem by optimally selecting the weights and basic functions of the wavelet network. However, most of these attempts are not suitable for online implementation. A recursive orthogonal least squares algorithm based on QR decomposition was proposed. Simulation results show that the structure of wavelet network can be designed and optimized by QR-ROLS algorithm effectively.

关 键 词：小波分析 神经网络 QR分解 RLS算法 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]