IJPAM: Volume 6, No. 4 (2003)

IDENTIFYING CHAOTIC REGIONS IN
THE RECURRENT NEURON PARAMETER SPACE

Athanasios Margaris, Nikos Kofidis, Manos Roumeliotis$^1$
Miltiadis Adamopoulos
Department of Applied Informatics,
Economics and Social Science
University of Macedonia
156 Egnatia Str., 540 06 Thessaloniki
P.O. Box 1591, GREECE
$^1$e-mail: manos@uom.gr


Abstract.The aim of this research is the experimental identification of chaotic regions in the parameter space of the recurrent neuron models. These models are used to describe the temporal behavior of a simple neural processing element with a self-interaction component that possesses chaotic characteristics in the discrete time domain. In the first step, the state equation of the system is described briefly, and the five coordinates of the state space points are identified. Then, the fixed points of the system are studied and sample bifurcation diagrams that reveal the chaotic features of the system are given. Finally, a complete experimental chaos detection procedure is presented and the identified chaotic regions in the parameter space associated with typical values of the system parameters are plotted.

Received: April 20, 2003

AMS Subject Classification: 70K50, 48F13

Key Words and Phrases: neuron models, chaotic regions, temporal behavior of a simple neural processing element, bifurcation diagrams, experimental chaos detection procedure

Source: International Journal of Pure and Applied Mathematics
ISSN: 1311-8080
Year: 2003
Volume: 6
Issue: 4