Attractor neural networks with excitatory neurons and fast inhibitory interneurons at low spike rates

Abstract

The mean-field formalism of attractor neural networks is extended to networks of interacting excitatory and inhibitory neurons in which the inhibitory neurons have a fast response to the excitatory neurons. A network of graded response neurons with unstructured inhibition has attractors with uniform low firing rates, describing the retrieval of single patterns, and pattern mixtures are suppressed by the network. In the case where there are extensively many patterns, the analysis is carried out in the replica symmetric approximation.