BISC Seminar Announcement , TUESDAY October 27th, 4-5pm, 310 Soda

Frank Hoffmann (fhoffman@cs.berkeley.edu)
Mon, 26 Oct 1998 06:44:57 +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
************************************************************

Soft Computing Applications: the Advent of Hybrid Systems


Speaker : Piero P. Bonissone

General Electric Corporate Research R&D
1 Research Circle, K1 5C32A, Schenectady, NY 12309, USA

Email: Bonissone@crd.ge.com

Date: TUESDAY, October 27th, 1998
Time: 4-5pm
Location : 310 Soda Hall

Abstract

As we attempt to solve real-world problems, we realize that they are
typically ill-defined systems, difficult to model and with large-scale
solution spaces. In these cases, precise models are impractical, too
expensive, or non-existent. The relevant available information is usually
in the form of empirical prior knowledge and input-output data
representing instances of the system's behavior. Therefore, we need
hybrid approximate reasoning systems capable of handling such imperfect
information.

Soft Computing (SC), a new field of Computer Sciences that deals with the
integration of problem-solving technologies such as Fuzzy Logic (FL),
Probabilistic Reasoning (PR), Neural Networks (NNs), and Genetic
Algorithms (GAs), provides us with a set of flexible computing tools to
perform these approximate reasoning and search tasks

We will focus on three real-world applications of SC that leverage the
synergism created by hybrid systems. Specifically we will describe:

1) The use of FL to control GAs and NNs parameters;
2) The use of GAs to tune FL controllers; and
3) The fusion of multiple models based on neural-fuzzy and fuzzy
case-based reasoning.

The first application illustrates the use of FL to implement a smart
algorithm-controller that allocates the algorithm's resources to improve
its convergence and performance. By tuning the algorithm parameters at
run-time, the fuzzy controller guide the search of a GA in the solution
space of an agile manufacturing problem. Another fuzzy controller is used
at training time to improve the convergence of NN algorithm used to
predict the paper web breakage in a paper mill.

The second application describes the use of GAs to tune a fuzzy controller
in situation where the evaluation of the fitness function is
computationally expensive. We describe the design and tuning of a
controller for enforcing compliance with a prescribed velocity profile for
a rail-based transportation system. We synthesize a fuzzy controller for
tracking the velocity profile, while providing a smooth ride and staying
within the prescribed speed limits. We use a genetic algorithm to tune
the fuzzy controller's performance by adjusting its parameters in a
sequential order of significance, resulting in a controller that is
superior to the manually designed one, and with only modest computational
effort.

The third application illustrates a fuzzy-rule based fusion process of
several prototype systems developed to estimate residential property
values for real estate transactions. One system uses Fuzzy Case-Based
Reasoning techniques to express preferences in determining similarities
between subject and comparable properties. These similarities guide the
selection and aggregation process, leading to the final property value
estimate. Fuzzy techniques are also used to generate a confidence value
qualifying such estimate. A second input to the fusion module is provided
by a modified version of ANFIS, a fuzzy-neural net trained on a subset of
cases from the case-base, that provides a similar estimate and confidence
value.

These applications exemplify the development of hybrid algorithms that are
superior to each of their underlying SC components and that provide us
with the better real-world problem solving tools.

**********************************************************************
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 NOT !! to bisc-group@diva.eecs.berkeley.edu
**********************************************************************

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

############################################################################ This message was posted through the fuzzy mailing list. (1) To subscribe to this mailing list, send a message body of "SUB FUZZY-MAIL myFirstName mySurname" to listproc@dbai.tuwien.ac.at (2) To unsubscribe from this mailing list, send a message body of "UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com" to listproc@dbai.tuwien.ac.at (3) To reach the human who maintains the list, send mail to fuzzy-owner@dbai.tuwien.ac.at (4) WWW access and other information on Fuzzy Sets and Logic see http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info (5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html