Chaotic dynamics of time-delay neural networks

A study is made of the nonlinear dynamical properties of a class of recurrent artificial neural networks with time delay that have been proposed in the literature to learn and recognize time-varying input patterns. Their nonlinear dynamics are studied to see if they have the potential to model nonlinear dynamical behavior of biological field potentials such as the electroencephalogram (EEG). It is shown that changing the value of the weights in a periodic fashion causes the output of time-delay neural networks to show a correlation dimension in the EEG range