Knowledge Acquisition Using Fuzzy Neural Network

L.C. Gabato (liza@pms.CS.Berkeley.EDU)
Tue, 5 Nov 1996 20:23:19 +0100


Knowledge Acquisition Using Fuzzy Neural Network

BISC Seminar (LIZA: please send to BISC-GROUP@diva and seminars@hera)

Takeshi Furuhashi

Berkeley Initiative in Soft
Computing Dept. of Information Electronics
387 Soda Hall Nagoya University
Computer Science Division Furo-cho, Chikusa-ku, Nagoya 464-01,
University of California and Japan
Berkeley, CA 94720-1776 USA Tel. +81-52-789-2792
Tel. +1-510-642-9827 Fax. +81-52-789-3166
Fax. +1-510-642-5775 Email

7 November 1996
310 Soda Hall


Description of input-output relationships of nonlinear systems from data is
one of many important knowledge acquistion problems. Artificial neural
networks (NN) is an effective tool for the identification of such models.
However, one disadvantage of NN modeling is that the knowledge acquired by
NN is hard to extract. Fuzzy Neural Networks (FNN) can use the Back
Propagation (BP) algorithm to identify and express input-output
relationships in the form of fuzzy rules, thus leading to possible knowledge
extraction by humans. In this talk, the speaker will present research on
applications of FNN to modeling nonlinear systems. Example applications to
be presented include modeling a steel making process, a weaving process, a
decision making process, a human's subjective comfort level, a sake (rice
wine) making process, etc.

Knowledge acquistion from data of nonlinear systems is a far reaching
problem and FNN does not solve all problems. Other hybrid techniques
researched by the speaker's group, such as hierachical fuzzy modeling
methods using FNN in combination with genetic algorithms (GA), will be

The contents of this talk will be:

1. Introduction of fuzzy neural network(FNN)
2. Fuzzy modeling using FNN
o Basic Idea
o Actual Application (a steel making process)
3. Further study
o Hierarchical fuzzy modeling
o Combination of FNN and GA

Back to Upcoming Seminars