Mike Mozer to speak in Colorado Machine Learning series

[Lorien Y. Pratt]

			      The spring, 1993
			  Colorado Machine Learning 
			      Colloquium Series
			       
				  presents:

			    Dr. Michael C. Mozer
			Department of Computer Science
		       University of Colorado, Boulder

	    Neural network approaches to formal language induction

			  Thursday, April 22, 1993
		  Room 110, Stratton Hall, on the CSM campus
				  5:30 pm

				 ABSTRACT

      I will describe several recent approaches to formal language
      induction using neural networks.  In contrast to 'connectoplasm'
      approaches that make use of generic network architectures, the work
      I'll advocate incorporates prior symbolic constraints to bias
      solutions learned by the network.  One network I'll describe, the
      demon architecture (Mozer & Das, 1993), learns symbol rewrite
      operations that allow it to construct parse trees for strings in
      LR(0) context-free grammars.  Another network, called SLUG (Mozer
      & Bachrach, 1991), processes regular languages and is based on an
      alternative and concise representation of some regular languages
      called an update graph.  Both networks are trained on negative as
      well as positive examples, and, because of their specialized
      symbolically-constrained architectures, generalize extremely well.
      Additionally, their weights can often be interpreted in terms of
      symbolic rules.


      Suggested background readings: SLUG: A Connectionist Architecture
      for Inferring the Structure of Finite-State Elements; Michael
      C.  Mozer and Jonathan Bachrach.  Machine Learning, volume
      7, 1991, pages 139-160; A Connectionist Symbol Manipulator
      that Discovers the Structure of Context-Free Languages.
      Michael C. Mozer and Sreerupa Das.  To appear in: C. L. Giles
      and S. J. Hanson and J. D. Cowan (eds.) Advances in Neural
      Information Processing Systems V.  San Mateo, CA: Morgan
      Kauffmann, 1993.  These readings are available on reserve at
      the Arthur Lakes Library at CSM.  Ask for the reserve package
      for MACS570, subject: Mozer.    Non-students can check
      materials out on reserve by providing a driver's license.

			    Open to the Public
	    Refreshments to be served at 5:00pm, prior to talk


    For more information (including a schedule of all talks in this
    series), contact:  Dr. L. Y. Pratt, CSM Dept. of Mathematical and
    Computer Sciences, lpratt at mines.colorado.edu, (303) 273-3878.  The
    speaker may be contacted at: mozer at dendrite.cs.colorado.edu.

			       Sponsored by:
     THE CSM DEPARTMENTS OF MATHEMATICAL AND COMPUTER SCIENCES, GEOPHYSICS,  
			DIVISION OF ENGINEERING, AND
				  CRIS*
*The Center for Robotics and Intelligent Systems at the Colorado School of Mines