Title :
A SOM Based Method for Classes Overlap Degree Evaluation
Author :
Lemeni, Ioan ; Tepus, Nicolae
Author_Institution :
Univ. of Craiova, Craiova
fDate :
July 27 2008-Aug. 1 2008
Abstract :
In a classification problem, the most difficult decision is to choose the artificial neural network (ANN) architecture that offers the best results. In this paper we present a method that permits to quickly evaluate the degree of overlapping between classes. Once we know this degree, we can easily choose the appropriate ANN architecture.
Keywords :
artificial intelligence; pattern classification; self-organising feature maps; SOM based method; artificial neural network; classification problem; Artificial neural networks; Computer architecture; Computer networks; Information technology; Multi-layer neural network; Multilayer perceptrons; Nearest neighbor searches; Neural networks; Testing; Training data; Classification; probability density estimation; self-organizing map (SOM);
Conference_Titel :
Computing in the Global Information Technology, 2008. ICCGI '08. The Third International Multi-Conference on
Conference_Location :
Athens
Print_ISBN :
978-0-7695-3275-2
Electronic_ISBN :
978-0-7695-3275-2
DOI :
10.1109/ICCGI.2008.55
