WIT Press


NN-modelling For Nonlinear Functions Approximations

Price

Free (open access)

Paper DOI

10.2495/HPC000231

Volume

23

Pages

10

Published

2000

Size

1,087 kb

Author(s)

A.N. Kastania & M.P. Bekakos

Abstract

NN-modelling for nonlinear functions approximations A.N. Kastania & M.P. Bekakos Department of Informatics, Athens University of Economics & Business, Abstract In this work the possibility to develop a pattern matching mathematical network architecture is investigated aiming to the assignment of each pattern to the correct mathematical modelling class. These mathematical modelling classes consist of eleven nonlinear models. An expert neural network for nonlinear functions approximations is built which can learn different mathematical modelling patterns, thus implying the ability to recognize different nonlinear mathematical modelling functions. The efficiency of the method is proved through the experimental results achieved. 1 Introduction MultiLayer Perceptions (MLPs) can be used to model nonlinear mappings. The main drawback of the MLPs is their rather slow trainability. The backpropagation of the error through the layers of the network takes a long time. An alternative model f

Keywords