Importance of parameter selection in classiffication systems using neural networks
- J. 1
- F. 2
-
1
Universidad de La Rioja
info
-
2
Universidad de Oviedo
info
ISSN: 1570-7946
Argitalpen urtea: 2000
Alea: 8
Zenbakia: C
Orrialdeak: 139-144
Mota: Artikulua
Beste argitalpen batzuk: Computer Aided Chemical Engineering
Laburpena
In this contribution we will try to set some "general" guidelines to select the criteria (and best suited parameters) based on many test raised on an industrial example, which is a heating regulation taking place without feedback in a power plant owned by a medium size electrical company located in the north of Spain. Also we will tray to analyze from an experimental point of view the ability of NN technology for modeling the process considering a low percentage of total patterns for training, just as a measure of its tolerance to the noise. © 2000 Elsevier B.V. All rights reserved.