An Associative Multivalued Recurrent Network

This paper shows that the multivalued recurrent neural model (MREM) can be used as an autoassociative memory (multivalued counterpart of Hopfield network). The architecture of the proposed family of networks is inspired from bipolar Hopfield’s neural network (BH). We have modified the function of energy of the bipolar Hopfield model by a new function of the outputs of neurons that we are naming function of similarity as it measures the similarity between the outputs of neurons. When binary neurons are considered and the function product is used as a function of similarity, then the proposed model is identical to that of Hopfield. We have studied a method to load a set of patterns into the network. That method corresponds to the Hopfield’s one when bipolar neurons are been considered. Finally we show that an augmented network avoids the storage of undesirable patterns into the network, as the well-known effect of loading the opposite pattern into the Hopfield’s network. For this new storage technics an expression that allows to set bounds to the capacity of the network is obtained.