THE EMERGENCE OF HIERARCHICAL CLUSTERED REPRESENTATIONS IN A HEBBIAN NEURAL-NETWORK MODEL THAT SIMULATES ASPECTS OF DEVELOPMENT IN THE NEOCORTEX

In this paper we show how clustered representations emerge within a si mple neural network which simulates some aspects of the development of the mammalian neocortex. Synapse death and stabilization, postulated to occur in the real neocortex are modelled. The representations that emerge under certain conditions show a spatially hierarchical coarse- and fine-grain cluster structure. In two sets of simulations where the relative lengths of excitory and inhibitory links are different we sh ow that the emergence of clustered representations depends upon change s to this parameter.