GINN (GENETIC INSIDE NEURAL-NETWORK) - TOWARDS A NONPARAMETRIC TRAINING

Authors
SANCHEZ MS SARABIA LA
Citation
Ms. Sanchez et La. Sarabia, GINN (GENETIC INSIDE NEURAL-NETWORK) - TOWARDS A NONPARAMETRIC TRAINING, Analytica chimica acta, 348(1-3), 1997, pp. 533-542
Citations number
21
Categorie Soggetti
Chemistry Analytical
Journal title
Analytica chimica acta → ACNP
ISSN journal
00032670
Volume
348
Issue
1-3
Year of publication
1997
Pages
533 - 542
Database
ISI
SICI code
0003-2670(1997)348:1-3<533:G(IN-T>2.0.ZU;2-N
Abstract
A genetic algorithm (steady state evolutionary algorithm without dupli cation) to train neural networks for classification problems is invest igated. The algorithm is based on direct optimization of frequencies o f both misclassifications and number of correct classifications with s pecial attention to the parameters known as sensibility and specificit y. This way, it is possible to find more information about the behavio ur of artificial neural networks on classification.