Representing properties in conceptual spaces towards effective learning within intelligent agents

Abstract

A conceptual spaces approach is a midway representation to bridge a gap between too detailed raw data from sensory receptors and too coarse symbolic knowledge representations. This paper provides a robust framework for representing properties and concepts that are core elements for many cognitive tasks within conceptual spaces. We use the Voronoi diagram and its dual Delaunay triangulation to define indeterminate property boundaries with lower and upper bounds. We compare and contrast this approach with some other polygonization methods.