An Empirical Study of Knowledge Representation and Learning within Conceptual Spaces for Intelligent Agents

This paper investigates the practicality and effectiveness of conceptual spaces as a framework for knowledge representation. We empirically compares and contrasts two popular quantitative lazy learning systems (nearest neighbor learning and prototype learning) within conceptual spaces and mere multidimensional feature spaces. Experimental results demonstrates conceptual spaces are superior to mere multidimensional feature spaces in concept learning and confirm the virtue of conceptual spaces.