🚀 go-pugleaf

                     CSGO Research Seminar
		     ---------------------	
CSGO, the Computer Science Graduate Organization, 
sponsors seminars to bring together students and 
researchers in the areas of Computer Science and 
Engineering, introducing the students to the 
researchers and providing a chance for the 
researchers to get to know the students.

Some of the talks during the semester will be given 
by senior Ph.D students.

There will be one seminar every week of approximately 
one hour duration followed by a question and answer 
session. 

The seminar is at 10:00 a.m. every Tuesday morning 
in SAL 324.
		  
Next week the seminar will be:
____________________________________________________

A Model of Primate Visual-Motor Conditional Learning
_____________________________________________________

Speaker : Andrew H. Fagg
	  Computer Science Department 
          University of Southern California
____________________________________________

		ABSTRACT
		--------

This talk will present work that I have done with Michael
Arbib on modeling the behavior and neural activity in 
premotor cortex of monkeys learning to pair an arbitrary 
visual stimulus with one of a set of previously-learned 
motor responses.  The network model is comprised of a 
large number of "motor selection columns." Reinforcement 
learning is used to tune the sensitivity of these columns 
so that they learn to recognize relevant visual features 
and produce support for the the appropriate motor 
response.  Although each column is allowed only limited 
information a set of columns are able to work together 
to learn a single visual-motor mapping.  

This model has been successful in reproducing a number of 
key behavioral and neurophysiological observations.  We 
will present a number of predictions for new experimental 
outcomes and will also show how recent data may be used 
to make a number of modifications to the model.

Finally, recent work will be discussed briefly which 
attempts to extend this model in two different directions:
  first, by adding a temporal component to the generation 
of output, and second, by beginning to think about how the 
monkeys might learn how to perform the task in the first 
place.  This work will then be related to learning in 
robots.

_________________________________________________________

			Date : Tuesday 10/13/92
			Time : 10:00 a.m.
			Place: SAL 324
			Host : CSGO
________________________________________________________