Analysis of CNN dynamics through spatio-temporal spectral techniques

It is shown that the main features of periodic attractors of cellular neural networks (CNNs) can be studied through a spatio-temporal spectral technique. Such a technique gives rise to a set of 3NM nonlinear transcendental equations for two-dimensional CNNs composed by NM cells. This set can be considerably simplified by exploiting the space-invariant structure of the CNN templates. Some results on the application of the spectral technique to one-dimensional CNNs are reported.