Přejít na obsah

Detail publikace

Citace

Šimandl, M. and Hering, P. : Nonlinear parameter estimation of neural network by Gaussian sum . Proceedings of International Carpathian Control Conference , p. 401-404, Technical University of Košice, Košice, 2003.

Abstrakt

Aim of the paper is to describe a new approach for the neural network parameter estimation of a nonlinear stochastic system. The mathematical model of the system created by the multilayer perceptron neural network is based on measured data exclusively assuming no or only diminutive knowledge about physics of the system. The paraters called weights in neural networks are estimated by nonlinear estimation method using Gaussian sum approach. The new optimization approach yields better parameter estimates comparing to commonly used methods.

Detail publikace

Název: Nonlinear parameter estimation of neural network by Gaussian sum
Autor: Šimandl, M. ; Hering, P.
Jazyk publikace: anglicky
Datum vydání: 26.5.2003
Rok vydání: 2003
Typ publikace: Stať ve sborníku
Název časopisu / knihy: Proceedings of International Carpathian Control Conference
Strana: 401 - 404
ISBN: 80-7099-509-2
Nakladatel: Technical University of Košice
Místo vydání: Košice
Datum: 26.5.2003 - 29.5.2003
/ /

Klíčová slova

system identification, nonlinear stochastic system, neural network parameter estimation, Gaussian sum, multilayer perceptron neural network

BibTeX

@INPROCEEDINGS{SimandlM_2003_Nonlinearparameter,
 author = {\v{S}imandl, M. and Hering, P.},
 title = {Nonlinear parameter estimation of neural network by Gaussian sum},
 year = {2003},
 publisher = {Technical University of Ko\v{s}ice},
 journal = {Proceedings of International Carpathian Control Conference },
 address = {Ko\v{s}ice},
 pages = {401-404},
 ISBN = {80-7099-509-2},
 url = {http://www.kky.zcu.cz/en/publications/SimandlM_2003_Nonlinearparameter},
}