辨识方法的计算效率(2):迭代算法

Computational efficiency of the identification methods. Part B: Iterative algorithms

讨论了最小二乘迭代辨识算法及其计算效率问题.最小二乘迭代算法由于涉及矩阵求逆运算,为减小计算量,提出了基于块矩阵求逆的最小二乘迭代辨识算法.基于块矩阵求逆的最小二乘迭代辨识算法不是一种新算法,只是从辨识算法的实现方式上降低计算负担,它与最小二乘迭代算法产生相同的参数估计,但计算量小.文中研究了伪线性回归系统、多元伪线性回归系统、多变量伪线性回归系统的最小二乘迭代辨识算法及其基于块矩阵求逆的最小二乘迭代算法.

This paper focuses on the computational efficiency of the least squares based iterative algorithms. The computational burdens of the least squares based iterative （LSI） algorithms are heavy due to computing large-size matrix inversion. In order to reduce the computational burdens, the block matrix inversion based LSI algorithms are presented. The proposed methods can reduce the computational cost through simplifying the implementation of the least squares based iterative algorithms, thus the estimation accuracies remain unchanged. The least squares based iterative algorithms and the block matrix inversion based LSI methods are studied for pseudo-linear regression systems, multivariate pseudo-linear regression systems and muhivariahle pseudo-linear systems.