Title :
Mean square convergence of multi-innovation forgetting gradient identification
Author :
Ding, Feng ; Ding, Tao
Author_Institution :
Dept. of Autom., Tsinghua Univ., Beijing, China
fDate :
6/23/1905 12:00:00 AM
Abstract :
A multi-innovation forgetting gradient identification method is studied; and its mean square convergence is analyzed by using stochastic process theory. The analysis indicates that the stationary data can improve accuracy of the parameter estimates, and that a way to choose the forgetting factor is obtained to minimize an upper bound of the parameter estimation error. The multi-innovation forgetting gradient algorithm is capable of reducing the effect of poor data in parameter estimation, having good robustness, and tracking time-varying parameters
Keywords :
convergence of numerical methods; gradient methods; minimisation; parameter estimation; signal processing; stochastic processes; error upper bound; forgetting factor; forgetting gradient method; identification method; mean square convergence; multi-innovation method; parameter estimation; signal processing; stationary data; stochastic process theory; time-varying parameters; tracking; Automation; Convergence; Least squares approximation; Parameter estimation; Robustness; Signal processing algorithms; Stochastic processes; Technological innovation; Time varying systems; Upper bound;
Conference_Titel :
Communications, Computers and signal Processing, 2001. PACRIM. 2001 IEEE Pacific Rim Conference on
Conference_Location :
Victoria, BC
Print_ISBN :
0-7803-7080-5
DOI :
10.1109/PACRIM.2001.953663
