RELATION BETWEEN TEMPLATE SPECTRUM AND CONVERGENCE OF DISCRETE-TIME CELLULAR NEURAL NETWORKS

A convergence criterion is proposed for reciprocal, discrete-time cellular neural networks, that exploits the peculiar interconnection structure of such networks. It is based on the Fourier spectrum of the discrete sequence representing the cloning template, and can be applied in the usual case where the neighbourhood is much smaller than the network. The corresponding design constraint, involving the template values, results in being less restrictive than those existing in the literature.