|
Vol: 52(66) No: 4 / December 2007 Regression Methods for Determination of a Multilayer Neural Network Architecture D. Ostafe “Stefan cel Mare” University of Suceava, Universitatii Street no.13, 720225 Suceava, Romania, phone: +4023021216147, e-mail: dumitruo@seap.usv.ro Gheorghe-Stefan Pentiuc “Stefan cel Mare” University of Suceava, Universitatii Street no.13, 720225 Suceava, Romania, e-mail: pentiuc@eed.usv.ro Stefan Holban Department of Computers, “Politehnica” University of Timisoara, Vasile Parvan Street no.2, Timisoara, Romania, phone: +40256404060, e-mail: stefan@aspc.cs.upt.ro Sorin Vlad “Stefan cel Mare” University of Suceava, Universitatii Street no.13, 720225 Suceava, Romania, e-mail: sorinv@seap.usv.ro Keywords: regression methods, neural networks, hidden layers, clustering methods Abstract The paper presents the technique used to determine the architecture of a neuronal network by means of the regression methods. Starting from the experimental results obtained by grouping the input data a regression model is used to determine the number of hidden layers and the number of neurons on each layer. References [1] Lakhmi Jain, Anna Maria Fanelli, Recent Advances in Artificial Neural Networks Desing and Applications, CRC Press , florida, USA, 2000. [2] Juan R. Rabunal, Julian Dorado, Artificial Neural Networks in Real-Life Applications, Idea Group Inc., London, United Kindom, 2006. [3] Mukesh Khare, S.M. Shiva Nagendra, Artificial Neural Networks in Vehicular Pollution Modelling, Springer-Verlag Berlin, 2000. |