Copenhagen Image and Signal Processing
Graduate School


Main page > Events > Talks > Genetic Fuzzy Systems

Main page
People
Events
     Courses
     Talks
     Workshops
Mailing list
Software
Presentations
Links
Contact

Genetic Fuzzy Systems


Lecturer: Frank Hoffman Center for Autonomous Systems CVAP
                        Numerical Analysis and Computing Science
                        Royal Institute of Technology, Stockholm 
                        http://www.nada.kth.se/~hoffmann/


Title: Genetic Fuzzy Systems

Place: Auditorium 5, HC ěrsted instituttet

Time: Friday October 25 at 9:15 a.m.

Abstract:

Within the realm of soft computing, neuro-fuzzy and genetic fuzzy
systems hybridise the approximate resaoning method of fuzzy logic with
the learning capabilities of neural networks and evolutionary
algorithms. This lecture focuses on genetic fuzzy systems and provides
an overview to genetic fuzzy rule based systems (GFRBS). In a GFRBS the
genetic process learns or tunes components of  the fuzzy knowledge base,
such as fuzzy membership functions and fuzzy rules. The lecture will
provide a brief overview on evolutionary algorithms and Mamdani and TSK
fuzzy rules. I will introduce the three major approaches for learning
rules, the Pittsburgh approach in which the chromosome represents an
entire rule set, the Michigan approach in which the population consists
of individual fuzzy rules and the iterative rule learning approach in
which the rule set is build in an incremental fashion. The key point of
a GFRBS is an evolutionary learning process that automates the knowledge
acquisition step in fuzzy system design. From the viewpoint of
optimisation the task of finding an appropriate knowledge base (KB) is
equivalent to parameterize the fuzzy KB (rules and membership functions)
and to find those parameter values that are optimal with respect to the
design criteria. Typical objectives in fuzzy modeling and decision
making are to minimize the modeling or classification error. In fuzzy
control the goal is usually to minimize a cost function such as the
integral of squared error. I will also discuss various genetic
representations of fuzzy rules and rule bases, based on fixed length,
variable length, position dependant and position independant coding
schemes.The lecture will review a few applications of genetic fuzzy
systems to modeling, classification and control.
-------------------------------------------------------------------------
The lecture is supported by Copenhagen Image and Signal Processing
Graduate School(CISP)http://www.imm.dtu.dk/cisp/

Later the same day, Frank will make an introduction to a discussion of
Evolutionary Robotics in N037, DIKU at 13:15

-- 
        Peter Johansen, Professor       
        DIKU, Universitetsparken 1, 
        2100 Copenhagen, Denmark
        http://www.diku.dk/users/peterjo/



$Id: 20021025Genetic.html,v 1.1 2002/10/23 14:29:57 fnielsen Exp $, © 2002 IMM DTU