Title :
Multi-innovation Stochastic Gradient Algorithm for Hammerstien Nonlinear ARX Systems
Author :
Liao, Yuwu ; Yu, Li ; Xiang, Lili
Author_Institution :
Sch. of Phys. & Electron. Eng., Xiangfan Univ., Xiangfan, China
Abstract :
Since the multi-innovation stochastic gradient (MISG) method can produce highly accurate parameter estimates for linear regression models, this paper extends the MISG method to nonlinear regression models and presents an MISG algorithm for Hammerstein nonlinear CAR (ARX) systems with memory less nonlinear blocks followed by controlled autoregressive models. The numerical results indicate that the proposed algorithm can effectively estimate the parameters of nonlinear systems.
Keywords :
autoregressive processes; gradient methods; nonlinear dynamical systems; parameter estimation; regression analysis; Hammerstein nonlinear CAR systems; Hammerstien nonlinear ARX systems; MISG method; controlled autoregressive models; linear dynamical blocks; linear regression models; memory less nonlinear blocks; multiinnovation stochastic gradient algorithm; parameter estimation; recursive identification; Computational modeling; Computers; Mathematical model; Parameter estimation; Signal processing algorithms; Stochastic processes; Hammerstein models; Parameter estimation; Recursive identification; Stochastic gradient;
Conference_Titel :
Genetic and Evolutionary Computing (ICGEC), 2010 Fourth International Conference on
Conference_Location :
Shenzhen
Print_ISBN :
978-1-4244-8891-9
Electronic_ISBN :
978-0-7695-4281-2
DOI :
10.1109/ICGEC.2010.220
