Paper available on factorial learning and EM

[Zoubin Ghahramani]

FTP-host: psyche.mit.edu
FTP-filename: /pub/zoubin/factorial.ps.Z

URL: ftp://psyche.mit.edu/pub/zoubin/factorial.ps.Z

This NIPS preprint is 8 pages long [300K compressed].


	       Factorial Learning and the EM Algorithm

			  Zoubin Ghahramani
	       Department of Brain & Cognitive Sciences
		Massachusetts Institute of Technology
			 Cambridge, MA 02139

Many real world learning problems are best characterized by an
interaction of multiple independent causes or factors. Discovering
such causal structure from the data is the focus of this paper.  Based
on Zemel and Hinton's cooperative vector quantizer (CVQ) architecture,
an unsupervised learning algorithm is derived from the
Expectation--Maximization (EM) framework. Due to the combinatorial
nature of the data generation process, the exact E-step is
computationally intractable.  Two alternative methods for computing
the E-step are proposed: Gibbs sampling and mean-field approximation,
and some promising empirical results are presented.

The paper will appear in G. Tesauro, D.S. Touretzky and T.K. Leen,
eds., "Advances in Neural Information Processing Systems 7", MIT
Press, Cambridge MA, 1995.