BISC Seminar Announcement, March 5th, 310 Soda, 4pm

Frank Hoffmann (fhoffman@cs.berkeley.edu)
Fri, 6 Mar 1998 16:43:19 +0100 (MET)

************************************************************
B I S C S e m i n a r A n n o u n c e m e n t
************************************************************

Issues in Automatic Identification of Fuzzy Rule-based Models

John Yen
Center for Fuzzy Logic, Robotics, and Intelligent Systems
Department of Computer Science
Texas A&M University
College Station, TX 77843-3112.
Email: yen@cs.tamu.edu

During the past decade, we have witnessed a rapid development
of various techniques for automatic identification (i.e., learning)
of fuzzy rule-based models. Most of these techniques use a set of
input-output data to train the model. However, two important
issues regarding fuzzy model identifications have not been fully
investigated: (1) the problem of model overfitting, and (2) the
issue of interpretability of Takagi-Sugeno-Kang (TSK) models. In
this talk, we will describe several techniques we have developed to
address these issues.
More specifically, we will discuss a model reduction technique
and an information-theoretic optimality criteria for fuzzy rule-based
models. These two techniques together have been shown to improve
the generalization capability of fuzzy models. The optimality
criteria formulates a trade-off between fitness of training data and model
simplicity, which is a fundamental principle underlying various general
theories regarding statistical modeling and inductive inference.
To address the interpretability issue of TSK model, we augmented the
global model fittness criteria with a local model fittness measure.
We will show empirical evidence that indicates that combining these two
criteria indeed resulted in TSK models with enhanced interpretability.

--------------------------------------------------------------------

Biography of John Yen

John Yen received his B.S. in electrical engineering from National
Taiwan University in 1980, and his Ph.D. in computer science from the
University of California, Berkeley in 1986. Dr. Yen is currently an
Associate Professor of the Department of Computer Science and the
Director of the Center for Fuzzy Logic, Robotics, and Intelligent
Systems at Texas A&M University. Before joining Texas A&M 1989, he
had been conducting AI research as a Research Scientist at Information
Sciences Institute of University of Southern California. His research
interests include intelligent agents, fuzzy logic, software engineering,
and pattern recognition. Dr. Yen is a member of the Board of Directors
of North American Fuzzy Information Processing Society (NAFIPS), the
Secretary of International Fuzzy Systems Association (IFSA), and
a member of IEEE NNC Technical Committee on Fuzzy Systems.
He has authored or coauthored over 100 technical journal and
conference papers. He has recently coauthored a textbook on fuzzy
logic (Fuzzy Logic: Intelligence, Control, and Information, Prentice
Hall, 1998). He is an Associate Editor of IEEE Transactions on Fuzzy Systems
and two other international journals on AI and fuzzy logic. Dr. Yen
received an NSF Young Investigator Award in 1992, the K. S. Fu Award from
NAFIPS in 1994, and the Dresser Industries Award from Texas A&M
University in 1995.

**********************************************************************
Please direct questions with regard to the contents of the talk
and request for papers to the speaker. To unsubscribe send email
to fhoffman@cs.berkeley.edu
**********************************************************************

Frank Hoffmann
---------------------------------------------------------------------------
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
---------------------------------------------------------------------------