小波神经网络在高维非线性系统辨识中的应用

在对高维系统的非线性辨识中 ,容易诱发维数灾难 ,而且由于数据分布的不均匀性和稀疏性 ,使得通常正交小波网络在应用时存在许多困难。本文利用非正交小波网络 ,按照文 [6 ]的方法训练和构造小波网络 ,针对其泛化能力不好的缺点 ,结合最小二乘算法提出了一种在线自适应调整网络权值的方法 ,获得了一类具有良好自适应能力的小波网络。

The Application of the Wavelet Networks in the Large Dimensionality Nonlinear System Identification

The nonlinear system modeling, especially of large dimension, often suffers from the curse of dimensionality. Furthermore, the application of the orthogonal wavelet network is very difficult for the non uniformly distributed and sparse training data. This paper uses the non orthogonal wavelet network as the modeling utility. First, we train the network with the techniques of regression analysis according to the algorithm in ; then, we present method self adapting and adjusting the weights online on the base of Least Squares Method, and get a kind of wavelet network with favorable self adapting ability.