Psychological concepts in a parallel system

Abstract : Many parallel information processing systems have ben proposed which are loosely based on architecture of the nervous system. Many of the testable predictions of these systems fall in the realm of cognitive science since they perform some 'computations' well and some poorly and have pronounced 'psychologies'. In such system, simple elements are connected to each other; the elements act like 'neurons' in that they sum excitation and inhibition from other elements. Information is represented as large state vectors of element activities, and it can be shown that simple 'synaptic' learing rules can serve to associate arbitrary state vectors using a matrix of connections. One version of this class of models has been shown,, in the past to form concepts and to perform classification computations. This paper discusses the general importance and structure of psychological concepts and describes three simulations of a simple neural model that attempts to reproduce a few of the important properties of human concept formation. Prototypes; Neural Models; Neural nets.