Two causes of instability of cellular neural networks
Patrick Thiran · 2003
Since a cellular neural network (CNN) is a nonlinear analog circuit, its solution may have a complex dynamical behavior, from complete stability to chaos. It is shown that this behavior may depend strongly on the boundary conditions set at the borders of the finite-sized CNN, the network being stable with some boundary conditions and unstable with others. Two kinds of completely unstable CNNs are distinguished: those that are unstable because of the boundary conditions and those that are unstable because of the template that defines them regardless of the boundary conditions.>