STUDY OF THE INFLUENCE OF NEURAL-NETWORK PARAMETERS ON THE PERFORMANCE-CHARACTERISTICS IN PATTERN-RECOGNITION

Authors
LI YW VANESPEN P
Citation
Yw. Li et P. Vanespen, STUDY OF THE INFLUENCE OF NEURAL-NETWORK PARAMETERS ON THE PERFORMANCE-CHARACTERISTICS IN PATTERN-RECOGNITION, Chemometrics and intelligent laboratory systems, 25(2), 1994, pp. 241-248
Citations number
8
Categorie Soggetti
Computer Application, Chemistry & Engineering","Instument & Instrumentation","Chemistry Analytical","Computer Science Artificial Intelligence","Robotics & Automatic Control
Journal title
Chemometrics and intelligent laboratory systems → ACNP
ISSN journal
01697439
Volume
25
Issue
2
Year of publication
1994
Pages
241 - 248
Database
ISI
SICI code
0169-7439(1994)25:2<241:SOTION>2.0.ZU;2-3
Abstract
In this paper, artificial neural networks (ANNs) with back-error propa gation were successfully used for the classification of complex fluori des (AB(m)F(n)) based on the transition emission of Eu(II). Network pa rameters and architecture were optimized. Effects of the transfer func tion, the scaling and learning mode and the range of the initial weigh ts on the performance of the network were studied. A recognition rate of 98.4% and a prediction rate of 96.3% showed better performance comp ared to conventional pattern recognition methods and previously report ed neural networks.