Forms of adapting patterns to Hopfield neural networks with larger number of nodes and higher storage capacity

This paper addresses forms of adapting patterns that have to be stored in a Hopfield neural network that contains a higher number of nodes than the dimension of such patterns we desire to store. With a brief introduction about the Hopfield network storage capacity subject, the paper presents the problem that appears when we have to adapt the length of the patterns to be stored, after a Hopfield network has its number of nodes increased. This increase in architecture size is frequently necessary in order to achieve a higher storage capacity and consequently a better recovery performance. Basically, three forms of adapting these stored patterns (and the probe vectors) are proposed. These three options are analyzed by experiments that use noisy versions of patterns as prompting vectors. Further, it discusses the particularities of each method proposed.