Title :
Variable parameter gradient estimation algorithm with local polynomial approximation
Author :
Xue, Yuncan ; Huang, Jiao ; Yang, Qiwen
Author_Institution :
Coll. of Comput. & Information Eng., Hohai Univ., Jiangsu, China
Abstract :
The trackability limitation of current gradient algorithm is discussed. A new algorithm, named variable parameter gradient estimation algorithm with local polynomial approximation is proposed. The concept of the error level is proposed. Selection criteria of the error level are given according to the min-max principle. Selection criteria of the order of the local polynomial are also given. This algorithm has good trackability and is particularly suitable for tracking time-varying systems. It is not more sensitive to the noise than the ordinary gradient estimation algorithms. The superior performance of proposed algorithm is verified via simulation studies.
Keywords :
gradient methods; parameter estimation; polynomial approximation; time-varying systems; local polynomial approximation; minmax principle; time-varying systems; variable parameter gradient estimation algorithm; Approximation algorithms; Computer errors; Educational institutions; Least squares approximation; Parameter estimation; Polynomials; Recursive estimation; Resonance light scattering; Steady-state; Time varying systems;
Conference_Titel :
Industrial Technology, 2003 IEEE International Conference on
Print_ISBN :
0-7803-7852-0
DOI :
10.1109/ICIT.2003.1290298