Towards Circuit Design with Auto-Associative Memories

In this paper, we discuss current and possible methods for synthesizing circuits using autoassociative neural networks. The ability of these networks to automatically converge to stable states when given an unknown input makes them suitable for storing and retrieving design information for a circuit. We demonstrate the use of the Hopfield network for realizing an amplifier and a filter. The limitations of the Hopfield network and the possible use of other neural network architectures are discussed in conclusion.