Advanced Tutorial on Learning DNF

[Dan Roth]

  Advanced Tutorial on the State of the Art in Learning DNF Rules
  ===============================================================

                       Sunday, July 10, 1994
                         Rutgers University
                      New Brunswick, New Jersey  

	 Held in Conjunction with the Eleventh International
       Conference on Machine Learning (ML94, July 11-13, 1994)
	  and the Seventh Annual Conference on Computational
	     Learning Theory (COLT94, July 12-15, 1994).


Learning DNF rules is one of the most important problems 
and widely-investigated areas of inductive learning from examples.
Despite its long-standing position in both the Machine Learning
and COLT communities, there has been little interaction between
them. This workshop aims to promote such interaction.

The COLT community has studied DNF extensively under its standard
learning models.  While the general problem is still one of the main
open problems in COLT, there have been many exciting developments
in recent years, and techniques for solving major subproblems 
have been developed.

Inductive learning of subclasses of DNF such as production rules,
decision trees and decision lists has been an active research topic
in the Machine Learning community for years, but theory has had almost
no impact on the experimentalists in machine learning working in this area.

The purpose of this workshop is to provide an opportunity for
cross-fertilization of ideas, by exposing each community to the other`s
ideas:  ML researchers to the frameworks, results and techniques
developed in COLT; the theoretical community to many problems that are
important from practical points of view, but are not currently
addressed by COLT, as well as to approaches that were shown to work in
practice but lack a formal analysis.

To achieve this goal the workshop is organized around a set
of invited talks, given by some of the prominent researchers 
in the field in both communities.
Our intention is to have as much discussion as possible during the
formal presentations.

The speakers are:

Nader Bshouty, University of Calgary, Canada
Learning via the Monotone Theory

Wray Buntine, NASA
Generating rule-based algorithms via graphical modeling

Tom Hancock, Siemens
Learning Subclasses of DNF from examples
     
Rob Holte, University of Ottawa, Canada
Empirical Analyses of Learning Systems

Jeff Jackson, Carnegie Mellon University
Learning DNF under the Uniform Distribution

Michael Kearns, AT&T Bell Labs
An Overview of Computational Learning Theory
Research on Decision Trees and DNF

Yishay Mansour, Tel-Aviv University, Israel
Learning boolean functions using the  Fourier Transform.

Cullen Schaffer, CUNY
Learning M-of-N and Related Concepts
	

PARTICIPATION

The Workshop is open to people who register to the COLT/ML
conference. We hope to attract researchers that are active 
in the area of DNF as well as the general COLT/ML audience.
 
WORKSHOP ORGANIZERS 

	Jason Catlett			Dan Roth
	
	AT&T Bell Laboratories 		Harvard University
	Murray Hill, NJ 07974   	Cambridge, MA 02138
	+1 908 582 4978			+1 617 495 5847 
	catlett at research.att.com	danr at das.harvard.edu