Associative memory neural networks for time series prediction

Author

Silver-Warner, S.J. ; Glover, R.J. ; Stonham, T.J.

Author_Institution

Dept. of Electr. & Electron. Eng., Brunel Univ., Uxbridge, UK

Volume

2

fYear

1996

fDate

13-16 Oct 1996

Firstpage

651

Abstract

Abus´ CMAC network is compared with a B-spline neural network for predicting the Mackey-Glass time series. The performance of the CMAC network is improved significantly through a combination of a smooth bell-shaped basis function and Parks and Militzer´s placement strategy. Despite the improvement of the CMAC network, the B-spline network gives significantly better performance. An example of trading modelling ability for a reduced number of parameters is given for the B-spline network

Keywords

cerebellar model arithmetic computers; content-addressable storage; splines (mathematics); time series; B-spline neural network; CMAC network; Mackey-Glass time series; associative memory neural networks; bell-shaped basis function; cerebellar model articulation controller; modelling ability; placement strategy; time series prediction; Approximation methods; Associative memory; Convergence; Function approximation; Hypercubes; Lattices; Multilayer perceptrons; Neural networks; Spline; Vectors;

fLanguage

English

Publisher

ieee

Conference_Titel

Electronics, Circuits, and Systems, 1996. ICECS '96., Proceedings of the Third IEEE International Conference on

Conference_Location

Rodos

Print_ISBN

0-7803-3650-X

Type

conf

DOI

10.1109/ICECS.1996.584446

Filename

584446