Implementation of cellular neural networks for heteroassociative and autoassociative memories

A design methodology of cellular neural networks (CNN) for heteroassociative and autoassociative memories is presented. A new synthesis procedure of continuous-time CNN for heteroassociative memories is developed, which assures global stability and robustness to the designed networks. A proper representation of discrete-time CNN characterized by multilevel output junctions is introduced to store memory vectors with b-bit length components. The suggested approach provides considerably simple network architectures suitable for VLSI implementation.