A novel encoding strategy for associative memory

Abstract

A novel encoding strategy for neural associative memory is presented in this paper. Unlike the conventional pointwise outer-product rule used in the Hopfield-type associative memories, the proposed encoding method computes the connection weight between two neurons by summing up not only the products of the corresponding two bits of all fundamental memories but also the products of their neighboring bits. Theoretical results concerning stability and attractivity are given. It is found both theoretically and experimentally that the proposed encoding scheme is an ideal approach for making the fundamental memories fLxed points and maximizing the storage capacity which can be many times of the current limits.