BISC Seminar, 18 Sept 97, 4-5pm

Frank Hoffmann (fhoffman@cs.berkeley.edu)
Thu, 18 Sep 1997 15:37:53 +0200 (MET DST)

B I S C S E M I N A R A N N O U N C E M E N T

==============================================
A Combined Approach to Fuzzy Reasoning
or: How to Cope with Complex Fuzzy Rule Bases
==============================================

Dr. Joachim Weisbrod
Forschungszentrum Informatik, Karlsruhe
GERMANY
weisbrod@fzi.de

(from 07/01/97 to 09/30/97 at ICSI, Berkeley)

Thursday, Sept 18, 1997
4:00-5:00pm
310 Soda Hall


Abstract:

Fuzzy Reasoning has proven to be a very successful mechanism,
especially in the field of fuzzy control. Nevertheless, fuzzy rule
bases are mostly very simple. Complex rule bases with several
layers of rules are hard to find. From our point of view, this
is mainly due to a lack in understanding of the theoretical
basics of fuzzy reasoning.

In this talk I am going to present a theoretical framework for
fuzzy reasoning. First, I examine the semantics of fuzzy
predicates like "x is A". From this study I derive two
complementary inference mechanisms, the well known possibilistic
approach on one hand, and the popular and successfull Mamdani
inference on the other hand.

The combination of both reasoning mechanisms promises to be
the key to coping with complex rule bases. By combining both
mechanisms, we are able to detect inconsistent and incomplete
information. Especially the detection and resolution of "gradual
contradictions" is crucial for infering reasonable consequents
over several layers of rules. Otherwise, local inconsistencies
accumulate with each additional inference step, finally leading
to useless answers.

---------------------------------------------------------------------------
Frank Hoffmann UC Berkeley
Computer Science Division Department of EECS
Email: fhoffman@cs.berkeley.edu phone: 1-510-642-8282
URL: http://http.cs.berkeley.edu/~fhoffman fax: 1-510-642-5775
---------------------------------------------------------------------------