Exponential Recurrent Associative Memories: Stability and Relative Capacity

Abstract

In this paper, relative capacity of a specific higher order Hopfield-type associative memory is considered. This model, which is known as exponential Hopfield neural network is suitable for hardware implementation and is not of a great computational cost. It is shown that, this modification of the Hopfield model significantly improves the storage capacity of the associative memory. We also classify the model via a stability measure, and study the effect of training the network with biased patterns on the stability