Relcon: A Tensor Model of Relational Categorization

Abstract

The Relcon algorithm models category formation using relational storage and retrieval mechanisms with a Tensor memory. Relcon settles on a category structure that matches prototypicality characteristics of human categories. A tensor intersection operation simulates the influence of context on category structure and on similarity. The results have implications for models of binding and representation, and provide a framework for answering the current call for relationbased representations for categories.