Heteroassociative memories via cellular neural networks

In this paper a synthesis procedure for heteroassociative memories using Cellular Neural Networks (CNNs) is presented. The suggested method, by assuring the condition of symmetry of the interconnection matrix, guarantees the complete stability of the designed network, besides providing that all the stored patterns correspond to asymptotically stable equilibrium points. Numerical examples are carried out to show the behaviour of the designed memory with respect to input perturbations. Moreover, the storage capacity and the presence of spurious equilibria have been investigated. © 1998 John Wiley & Sons, Ltd.