The race to the attractor model for classifying objects

The human brain is exceptionally good at classifying objects quickly and reliably. We can recognize familiar faces even when seen from different angles, despite irrelevant clutter such as jewellery, sunglasses, new hair styles, etc. Over the years, scientists have tried to duplicate this remarkable ability using neural networks models, but without much success. In this article, we examine some of the key characteristics that make the brain such an efficient tool for object recognition. We propose mechanisms through which these characteristics can be modeled. Then, we describe a novel approach to simulating object recognition with artificial neural networks. We used the recognition process for "Uncle Brian" to demonstrate how this model captures these key characteristics and achieves reliable classification.