A neurocomputational model of early psychosis

Abstract

A neurocomputation model for early psychosis is implemented with an associative network using the Hebbian learning algorithm. The model emphasizes on longitudinal evolution of the weight matrix from the pre-psychotic phase through the psychotic phase to the recovery phase. Enhanced activity-dependent synaptic plasticity is postulated to underlie formation of spurious associative connections during psychosis. This parsimonious longitudinal model is capable of addressing the presence of low- grade psychotic symptoms in the prepsychotic phase, the formation of spurious associations during psychosis, as well as the persistence of some of these associations into the recovery process with encapsulation of psychotic features.