Proceedings of the ISCIE International Symposium on Stochastic Systems Theory and its Applications
Online ISSN : 2188-4749
Print ISSN : 2188-4730
The 44th ISCIE International Symposium on Stochastic Systems Theory and Its Applications (Nov. 2012, Tokyo)
Multi-layered GMDH-type Neural Network Algorithm Using Principal Component-Regression Analysis and PSS Criterion
Tadashi KondoJunji UenoShoichiro Takao
Author information
JOURNAL FREE ACCESS

2013 Volume 2013 Pages 273-278

Details
Abstract
In this study, a hybrid Group Method of Data Handling (GMDH)-type neural network algorithm using principal component-regression analysis is proposed and applied to the nonlinear system identification. The architectures of the GMDH-type neural networks are automatically organized using heuristic self-organization method which is a kind of evolutionary computation.In the heuristic self-organization method, many nonlinear combinations of the input variables are generated and new neurons are constructed from these combinations. Only desirable neurons which fit the characteristics of the nonlinear system, are selected and these neurons are combined again in next layer. These procedures are iterated and a multi-layered neural network architectures are automatically organized. In the GMDH-type neural network, the multi-colinearity of the variables generates and the prediction output values of the neural network become unstable. In this study, the principal component-regression analysis is used for estimating the parameters of the neurons and stable and accurate multi-layered architectures of the GMDH-type neural networks are organized using the heuristic self-organization.
Content from these authors
© 2013 ISCIE Symposium on Stochastic Systems Theory and Its Applications
Previous article Next article
feedback
Top