Tuning Rules by Cases

	      Yoshio Nakatani	                  David Israel
	Industrial Electronics & Systems Lab.	       CSLI
	    Mitsubishi Electric Corp.	       Stanford University
	  Amagasaki, Hyogo 661, Japan	   Stanford, CA 94305-4115, USA


Abstract. A new method is proposed for tuning rules by cases, especially in domains in
which precise and exceptionless rules are known to be unavailable. When the result of
execution of a rule is not satisfactory, the system stores the name of the executed rule, the
conditions under which the rule was executed, the evaluation of the execution, the attributes 
and values to be modified, and hypothesized alternatives, as a case. The next
time the rule is to be executed under the same conditions, the relevant attributes and
values are temporarily modified, by replacement by their hypothesized alternatives. After 
a certain number of such experiments, the maintainer of the system can reconstruct
the whole rule base by referring to the stored cases. This methodology is implemented as
a system, A LA CARTE, in the domain of cooking.

References
1.	Branting, L. K. and Porter, B.W.: Rules and Precedents as Complementary Warrants, 
Proc. of AAAI-91 (1991).

2.	Golding, A. R. and Rosenbloom, P.S.: Improving Rule-Based Systems through
Case-Based Reasoning, Proc. of AAAI-91 (1991).

3.	Hammond, K. J.: Case-Based Planning: Viewing Planning as a Memory Task. Academic 
Press 1989, San Diego.

4.	Kolodner, J. L. : Towards an Understanding of the Role of Experience in the Evolution 
from Novice to Expert, Int. J. Man-Machine Studies, 19 (1983).

5.	Kolodner, J. L.: An Introduction to Case-Based Reasoning, Artificial Intelligence
Review, 6(1992).

6.	Lave, J. : Cognition in Practice: Mind, Mathematics and Culture in Everyday Life,
Cambridge, UK, Cambridge University Press (1988).

7.	Minsky, M.: Logical Versus Analogical or Symbolic Versus Connectionist or Neat
Versus Scruffy, AI Magazine, SUMMER (1991).

8.	Miyake, N. and Hatano, G.: Socio-cultural Constraints and Beyond, Advances in
Japanese Cognitive Science, Vol. 4 (1991) in Japanese.

9.	Nakatani, Y. and Israel, D. : An Architecture for Tuning Rules by Cases, Report No.
CSLI-92-173, CSLI, Stanford University (1992).
