A sequential dynamic heteroassociative memory for multistep pattern recognition and one-to-many association

Bidirectional associative memories (BAMs) have been widely used for auto and heteroassociative learning. However, few research efforts have addressed the issue of multistep vector pattern recognition. We propose a model that can perform multi step pattern recognition without the need for a special learning algorithm, and with the capacity to learn more than two pattern series in the training set. The model can also learn pattern series of different lengths and, contrarily to previous models, the stimuli can be composed of gray-level images. The paper also shows that by adding an extra autoassociative layer, the model can accomplish one-to-many association, a task that was exclusive to feedforward networks with context units and error backpropagation learning.

I. Introduction

Unsupervised recurrent autoassociative memories have received much attention since Hopfield's network [1]. They are well-suited to learn and recall correlated patterns, and their dynamic properties enable them to exhibit attractor behavior, an essential requirement for noisy input recall and pattern completion processes. The bidirectional associative memory (BAM) is a generalization of the Hopfield network in the case of supervised learning [2], [3]. In particular, the BAM network retains the dynamic feedback feature of the former, thus gaining the potential to correctly classify noisy or partial inputs. The models reported in the literature have been studied mainly in the context of learning one-to-one association.