递推计算灵敏度的初值正交计算法

AN ORTHOGONAL COMPUTATION METHOD OF INITIAL VALUE FOR RECURSIVE COMPUTING SENSITIVITIES

本文对连续-离散系统的极大似然估计方法提出了一种灵敏度初值的正交计算法,它是对具有递推计算灵敏度的修正牛顿-拉夫森算法的改进。文中通过借用正交试验的概念和正交表的性质,解决了原有算法的初始矩阵求逆所存在的困难。

This paper presents an orthogonal computation method of initial value of sensitivity for maximum likelihood estimation method to he continuous-discrete system. This method is an improvement to the modified Newton-Raphson scheme with estimated sensitivities (MNRES) method. The difficulty to invert the initial matrix in MNRES method is solved by the concept of orthogonal test and orthogonal table