Systolic-based hardware realisation of Hopfield neural network

Abstract

The paper proposed a new methodology for Hopfield neural network simulation based on ring systolic array structure. The methodology is related to training methods: Hebbian and Delta rules as well as to retrieving phase. The discussion is focused on operations which are realized during the following steps of each algorithm and the data transferred among the calculation units. The efficiency is discussed based on the modified set of parameters defined by S.Y. Kung.