Rumelhart, D.E., and Todd, P.M. (1993). Learning and connectionist representations. In D.E. Meyer and S. Kornblum (Eds.), Attention and performance XIV (pp. 3-30). Cambridge, MA: MIT Press/Bradford Books.

Abstract

Connectionist modeling is undergoing a renaissance. As the merits of brain-style computation (Rumelhart, 1990) have become apparent, a bewildering variety of connectionist applications have cropped up throughout the cognitive sciences and engineering (for instance, see Lippmann, Moody, and Touretzky, 1991). One of the central issues in all of these models is the representation of knowledge in the connectionist network. Getting a coherent picture of "what goes on" inside a network as it develops, manipulates, and alters the representation of the knowledge it processes is vital for our understanding of connectionist information processing, and likely for our understanding of the minds these systems model. In this paper we explore the sorts of representations that connectionist systems employ, and the crucial role learning plays in constructing them.



   Back to main page.
   View publications categorized by topic.
   View publication abstracts searchable by keywords