recurrent network of perceptrons with three state synapsesachieves competitive classification on real inputs

We describe an attractor network of binary perceptrons receiving inputs from a retinotopicvisual feature layer. Each class is represented by a random subpopulation of the attractor layer,which is turned on in a supervised manner during learning of the feed forward connections. Theseare discrete three state synapses and are updated based on a simple field dependent Hebbian rule.For testing, the attractor layer is initialized by the feedforward inputs and then undergoes asynchronousrandom updating until convergence to a stable state. Classification is indicated by thesub-population that is persistently activated. The contribution of this paper is twofold. First,this is the first example of competitive classification rates of real data being achieved throughrecurrent dynamics in the attractor layer, which is only stable if recurrent inhibition is introduced.Second, we demonstrate that employing three state synapses with feedforward inhibition is essentialfor achieving the competitive classification rates due to the ability to effectively employboth positive and negative informative features.