DISCOVERING PREDICTABLE CLASSIFICATIONS
(Neural Computation, 5(4):625-635, 1993)

Jürgen Schmidhuber¹
Department of Computer Science
University of Colorado
Boulder, CO 80309, USA

Daniel Prelinger
Institut für Informatik
Technische Universität München

Abstract:

Prediction problems are among the most common learning problems for neural networks (e.g. in the context of time series prediction, control, etc.). With many such problems, however, perfect prediction is inherently impossible. For such cases we present novel unsupervised systems that learn to classify patterns such that the classifications are predictable while still being as specific as possible. The approach can be related to the IMAX method of Hinton, Becker and Zemel (1989, 1991). Experiments include a binary stereo task proposed by Becker and Hinton, which can be solved more readily by our system.