MSIMS@RUTGERS.ARPA (04/03/84)
From: Michael Sims <MSIMS@RUTGERS.ARPA>
[Forwarded from the Rutgers bboard by Laws@SRI-AI.]
Machine learning brown bag seminar
Title: Automating Shifts of Representation
Speaker: P. J. Riddle
Date: Wednesday, April 11, 1984, 12:00-1:30
Location: Hill Center, Room 254
My thesis research deals with automatically shifting from one
knowledge representation of a certain problem to another representation
which is more efficient for the problem class to which that problem
belongs. I believe that "...changes of representation are not isolated
'eureka' phenomena but rather can be decomposed into sequences of
relatively minor representation shifts". I am attempting to discover
primitive representation shifts and techniques for automating them. To
achieve this goal I am attempting to define and automate all the
primitive representation shifts explored in the Missionaries &
Cannibals (M&C) problem. The main types of representation shifts which
I have already identified are: forming macromoves, removing irrelevant
information, and removing redundant information. Initially I have
concentrated on a technique for automatically acquiring macromoves.
Macromoves succeed in shifting the problem space to a higher level of
abstraction. Assuming that the macromoves are appropriate for this
problem class, this will make the problem solver much more efficient
for subsequent problems in this problem class.