|
(Back to top of
IJCNN I)
IJCNN I
"Neural Networks That Actually Work In Prediction and Decision/Control:
Common Misconceptions Versus Real-World Success"
Paul J. Werbos
NSF
Abstract: Neural nets are used
in a growing variety of real-world applications, in a growing number of
areas (including the modeling of natural intelligence). But this has been
a mixed blessing, as various subcultures have not kept up with each other.
For example, at http://www.eas.asu.edu/~nsfadp/ applications of adaptive
dynamic programming (aka advanced reinforcement learning) are reported
which beat the previous best performance in electric power grid control,
aircraft control, in large-scale logistics, etc. -- but best performance
requires integrating concepts from many subcultures. Yet in some
subcultures, people are still using "mappings from sensory to motor
coordinates" (direct adaptive control), whose limits were clear years ago.
Some are using indirect methods inspired by control theory with well-known
stability theorems -- theorems whose conditions are rarely satisfied, and
methods which are easily improved upon. Similarly, in prediction, some
subcultures falsely believe that recurrent networks are hard to train and
of marginal benefit, even as people in industry use them routinely, and
report performance better than extended Kalman filters and equal (at lower
cost) to elaborate particle filter methods. The foundations of learning
also bear on the practical choices. This tutorial will try to offer a kind
of practical roadmap of these issues, and suggest how it points towards
future functionality in tasks as large-scale as what brains can handle.
Vita: Paul J. Werbos is best known as the original inventor of
backpropagation, as part of his Harvard PhD thesis, which was reprinted in
full in his book the Roots of Backpropagation, Wiley 1994, along with his
classic 1990 tutorial on backpropagation through time for Proc IEEE. Also,
his 1967 article in Cybernetica first proposed the idea of approximating
dynamic programming as a way to improve reinforcement learning, elaborated
in many later papers. He was one of the three original two-year presidents
of the International Neural Network Society, and winner of the IEEE
Pioneer Award. He is Program Director for Control, Networks and
Computational Intelligence at NSF, which actively seeks more proposals in
this area. He has also been active in many cross-cutting funding
initiatives; for example, he serves on the Working Group for energy
storage and distribution of the interagency Climate Change Technology
Program, and coordinated the NASA-NSF-EPRI solicitation on space solar
power (NSF 02-098). He is also on the Planning Committee of the Millennium
Project of the United Nations University (http://millennium- project.org),
and has published a few papers on quantum foundations and technology (see
arXiv.org, physics and nonlinear systems). He also has two degrees in
economics from Harvard and the London School of Economics. He is on the
governing boards of INNS and of the IEEE Industrial Electronics Society.
(Back to top of
IJCNN I)
"Kalman Filter Training of Neural Networks: Methodology and Applications"
Danil V. Prokhorov
Ford Research and Advanced Engineering
Abstract:
We would like to share with you our
experience with neural network training acquired over many years of
application-driven studies at Ford Motor Company. Our main focus is on
training recurrent neural networks (RNN) using a set of powerful
techniques based on the Kalman filter method. RNN considered here are
very flexible discrete-time computational structures often called
time-lagged RNN. (For example, recurrent multilayer perceptrons (RMLP)
are a special case of such RNN.) The trick is how to train them. This
tutorial is for anybody who wants to learn arguably the most effective
training methodology for RNN.
Our studies are carried out both with models and
using real data gathered from physical systems. We intend to cover
backpropagation through time (BPTT), extended Kalman filter (EKF),
multi-stream EKF and other EKF extensions, nonlinear Kalman filter and
joint estimation of states and weights, followed by examples.
Vita: Danil Prokhorov began his technical
career in St. Petersburg, Russia. He received his Diploma in Robotics in
1992. In late 1993 he moved to the U.S. to study for a Ph.D. in
Electrical Engineering with emphasis on artificial neural networks. He is
one of the early students and active proponents of the Kalman Filter NN
Training. Upon graduation in late 1997, he joined Ford Research
Laboratory in Detroit Metropolitan area to carry out machine learning
research and development.
(Back to top of
IJCNN I)
"Neural
Control Systems"
Bernard Widrow
Stanford University
Abstract: This tutorial will
describe a variety of adaptive control algorithms for linear and nonlinear
systems. These topics include adaptive inverse control, self-learning
neural control systems, and neurointerfaces that allow human operators,
providing high-level command inputs, to control complex nonlinear systems.
At present, the control of a dynamic system (the "plant") is generally
done by means of feedback. Adaptive inverse control is an alternative
approach that uses adaptive filtering to achieve feedforward control.
Precision is attained because of the feedback incorporated in the adaptive
learning. Adaptive inverse control places an adaptive filter whose
transfer function converges to the inverse or reciprocal of that of the
plant in cascade with it. If the plant is minimum-phase, an inverse is
easily obtained. If the plant is nonminimum-phase, a delayed inverse can
be obtained. Plant disturbance can be optimally controlled by a special
circuit that obtains the disturbance at the plant output, filters it, and
feeds it back into the plant input. The circuit works in such a way that
the feedback does not alter the plant
dynamic response. So, optimal disturbance control and control of dynamic
response can be accomplished independently.
Self-learning neural controls apply to nonlinear plants that can be
started with arbitrary initial conditions, and their neural controllers
learn to bring the plants to desired final conditions.
Learning is accomplished by "backpropagation through time." Applications are
shown to the "truck-backer," backing a truck with one or two trailers to a
loading platform, to the control of a construction crane, and to robotics.
Neurointerfaces are trainable filters that serve as couplers between human
operators and nonlinear systems or plants to be controlled or
directed. The purpose of the coupler is to ease the task of the human
controller. The equations of a given plant are assumed to be known or
at least approximately known. If the plant is unstable, it must first be
stabilized by feedback. Using the plant equations, off-line
automatic learning algorithms are developed for training the weights of
the neurointerface. If the plant is subject to disturbance, an adaptive
disturbance canceller is used to minimize the effect. The neurointerface
can be adapted to be an approximate inverse of the plant, so that when it
is cascaded with the plant, the overall plant response closely approximates the human command input, making it easy for the human
operator to control difficult-to-control plants.
Vita: Bernard Widrow received the S.B., S.M., and Sc.D. degrees in
Electrical Engineering from the Massachusetts Institute of Technology in
1951, 1953, and 1956, respectively. He joined the MIT faculty and taught
there from 1956 to 1959. In 1959, he joined the faculty of Stanford
University, where he is currently Professor of Electrical Engineering.
He began research on adaptive filters, learning processes, and artificial
neural models in 1957. Together with M.E. Hoff, Jr., his first doctoral
student at Stanford, he invented with LMS algorithm on in the Autumn of
1959. Today, this is the world's most widely used learning algorithm. He
has continued working on adaptive signal processing, adaptive controls,
and neural networks since that time.
Dr. Widrow is a Life Fellow of the IEEE and a Fellow of AAAS. He received
the IEEE Centennial Medal in 1984, the IEEE Alexander Graham Bell Medal in
1986, the IEEE Signal Processing Society Medal in 1986, the IEEE Neural
Networks Pioneer Medal in 1991, the IEEE Millennium Medal in 2000, and the
Benjamin Franklin Medal for Engineering from the Franklin Institute of
Philadelphia in 2001. He was inducted into the National Academy of
Engineering in 1995 and into the Silicon Valley Engineering Council Hall
of Fame in 1999.
Dr. Widrow is a past president and currently a member of the Governing
Board of the International Neural Network Society. He is a member of the
AdCom of the IEEE Computational Intelligence Society. He is
associate editor of several journals and is the author of over 100
technical papers and 18 patents. He is co-author of "Adaptive Signal
Processing" and "Adaptive Inverse Control," both Prentice-Hall books. A
new book, "Quantization Noise," is in preparation.
(Back to top of IJCNN I)
"Feature extraction in Computational Intelligence"
Evangelia Micheli-Tzanakou, PhD
Biomedical Engineering
Rutgers Univ.
Abstract: One of the major
problems a researcher faces is what is learned from data obtained by
various methods and different techniques. This tutorial will discuss and
compare topics such as:
Statistical Advances and Challenges,
as well as
Feature Extraction in Computational Intelligence methodologies.
Often a simple model describes the data well, simply because the S/N ratio
is too small for detection of more complex structures-which for example is
the case with medical data involving human subjects. One has a lot of
variability both in intra- and inter-sets of data.
Some important Simple Tools that have been used for a long time are:
Linear Regression, Discriminant analysis, Principal Component Analysis
etc.
In all of these, the size of the data set matters. Huge data sets create
memory problems. The question is how do we handle different data types and
how do we handle them? What if the data are correlated? What if we have
complex data structures?
In this tutorial you will learn more about Computational Intelligence, how
to get to know your data and how to do Feature Extraction.
Some examples of "features" will be given and different feature extraction
methods will be discussed.
Vita: Evangelia Micheli-Tzanakou Dr. Tzanakou is Professor II and
director of the Computational Intelligence Laboratories (2000-present);
Chair of Biomedical Engineering Dept at Rutgers University
(1990-2000);established the BME Undergraduate curriculum in 1999. Has an
active research group and has published over 250 scientific papers.
Author/co-author of 2 books. Has graduated 37 MS and PhDs.
Fellowships: IEEE Fellow (1992);Founding Fellow, American Institute for
Medical & Biological Eng. (AIMBE), 1993; Member of Sigma Xi and Eta Kappa
Nu; Listed in American Men and Women in Science; Honorary Member of the
British Brain Research Association; Honorary Member of the European Brain
and Behavior Society.
Awards: NJ Women of Achievement Award, 1995; Featured in the “Notable
Twentieth Century Scientists, 1994; Achievement Award Recipient, Society
of Women Engineers, 1992; Outstanding IEEE Branch Advisor/Counselor Award,
1985 ; cited as a Pioneer in the IEEE Web page (http://www.ieee.org/innovators).
Editorships: Editorial Board, IEEE Trans. On Nano-BioScience
(2002-present); Editorial Board Biomedical Engineering On Line
(2001-present ); Associate Editor, IEEE Transactions on Neural Networks
(2000-present); Book Series Editor in Biomedical Engineering,
Plenum/Klewer (1999-present); Book Series Editor, Biomedical Systems, IES
Book series, CRC Press (1997-present); Editorial Board, IEEE Trans.
Biomedical Information Tech.(1997-2001); Editorial Board, International J.
On Advanced Computational Intelligence (1997-1999); Associate Editor, IEEE
Transactions on Neural Networks (1989-1992).
(Back to top of IJCNN I)
|
|
(Back to top of
IJCNN II)
IJCNN II
"Autonomous Mental Development: A New Frontier for
Computational Intelligence"
Juyang Weng
Professor, Embodied Intelligence Laboratory
Department of Computer Science and Engineering
Michigan State University
Abstract:
What is autonomous mental development (AMD)?
Why is it a new frontier for computational intelligence? Why do
neural networks, genetic computation and fuzzy computation all converge on
the new ground of mental development? What do recent results in
neural science imply for computational intelligence? Is human brain
a signal processor or a symbolic processor? Why can human
programmers write excellent computer chess playing programs but vision,
speech and language are still grand challenges for computers?
Can a machine develop these capabilities autonomously? How can
we enable a machine to learn autonomously online in real time directly
from the physical world of complex human environment?
We will address these fundamental
questions in this tutorial. We will systematically study a concept
called task muddiness which explains why traditional approaches to machine
intelligence “have run out of steam” and why AMD is necessary for machines
to deal with muddy tasks. Computational modeling of autonomous
mental development is a new field that has drawn increasing attention from
fields such as robotics, multimedia signal processing, vision, speech,
machine intelligence, neuroscience and psychology. In this tutorial,
we will present computational models that enable robots to develop its
mental skills autonomously through online, real-time interactions with its
environment. Such robots are called developmental robots and the
program that enables AMD by robots is called developmental program.
AMD requires that the internal representation and some architecture of the
robot be generated automatically and incrementally through the
developmental process. The goal of this new research
field is to enable humans to “raise” developmental robots “mentally”
through online, interactive “robot sitting” and “robot classes.”
Computational models of AMD and verification of the models on robots can
potentially cast new light on the way human brain works. This
tutorial will describe several robots at MSU and elsewhere as early
prototypes of such a new kind of robot along with some video
demonstrations. The potential social and economical impact of
developmental robots will also be discussed.
Tutorial topics:
-
Muddiness of tasks
-
Overview of approaches ---
knowledge-based, learning-based, behavior-based, evolutional and the new
developmental approach
-
Human mental development, results
from neuroscience and developmental psychology
-
Review of animal learning theories
-
Supervised, reinforcement and
communicative learning
-
Architectures for automatic mental
development
-
Sensory mapping: representation,
development and computation
-
Cognitive mapping: representation,
development and computation
-
Motor mapping: representation,
development and computation
-
Value system
-
Integration of mental capabilities:
audition, touch, language, reasoning, decision making, planning, object
manipulation and navigation
-
Thinking by a developmental robot
-
Examples of developmental robot
-
Theoretical completeness and
performance metrics
-
Applications and the future of
developmental robots
Length: 2 hours
Prerequisites: General
programming experience, basic knowledge about vector and matrix
operations. Students in neuroscience and psychology can also
understand most of the material.
Primary audience: researchers in
signal processing, image processing, computer vision, pattern recognition,
speech recognition, language processing, robotics, human-machine
interface, artificial intelligence and new-generation computer
architecture.
Secondary audience: researchers
in neuroscience, psychology, philosophy, and government policy makers for
science and technology.
Handout: Tutorial material will be
provided by the host institution.
Vita:
Juyang Weng is a professor at the Department of Computer
Science and Engineering, Michigan State University, East Lansing,
Michigan, USA. His research interests include computer vision,
speech recognition, human-machine multimodal interface using vision,
audition, speech, gesture and actions, and intelligent robots. He is
the author or coauthor of over one hundred research articles and book
chapters. He is a coauthor (with T. S. Huang and N. Ahuja) of
the book Motion and Structure from Image Sequences (Springer-Verlag,
1993). He is an editor-in-chief of International Journal of
Humanoid Robotics, an associate editor of IEEE Trans. on Pattern
Recognition and Machine Intelligence. He was an associate editor of
IEEE Trans. on Image Processing (1994-1997), a program co-chair of the
NSF/DARPA Workshop on Development and Learning (WDL), held April, 5-7,
2000 at Michigan State University (MSU), East Lansing, MI
(http://www.cse.msu.edu/dl/), and a program co-chair of the International
Conference on Development and Learning 2002 (ICDL’02), held at
Massachusetts Institute of Technology, Cambridge, MA, June 12- 15, 2002
(http://www.egr.msu.edu/icdl02/). An article “Autonomous Mental
Development by Robots and Animals,” that he co-authored appeared in
Science, the issue of Jan. 26, 2001. He initiated and supervised the
SAIL (Self-organizing Autonomous Incremental Learner) and Dav projects, in
which he and his coworkers have designed and custom built their SAIL and
Dav robots for research on autonomous mental development. More
detail is available on line at http://www.cse.msu.edu/~weng/.
(Back to top of
IJCNN II)
"Cellular Visual Microprocessors:
Theory, Implementation and Applications"
Csaba Rekeczky and Ákos Zarándy
Hungarian Academy of Sciences
Abstract: The aim of the session is to present the latest results
on theory, design and applications of cellular visual microprocessors and
topographic array computing on data flows. Cellular Neural Network (CNN)
based architectures will be put into focus with embedded optical sensors,
i.e. neural network based visual array
microprocessors. Overviews will be given on theory and design;
neurobiological motivations and models; different analog and mixed-signal
chip implementations and some key application areas will be highlighted.
By emphasizing concrete chip implementations in the CNN field, we hope
this tutorial will provide
attendees unfamiliar with recent advances an easily accessible way to
appreciate the potential impact of the work in this area, as well as
experienced researchers with an update of the latest results. This session
could also be viewed as a forum to encourage discussion about other neural
network hardware realizations having a particular set of applications in
mind.
Vitas:
Csaba Rekeczky was born in 1968. He received the M.S. degree in
electrical engineering from the Technical University of Budapest in 1993.
After graduation he joined the Neuromorphic Information Technology
interdisciplinary postgraduate program and continued his studies at the
Analogic and Neural Computing Systems Laboratory of the Computer and
Automation Research Institute of the Hungarian Academy of Sciences. During
this period he completed research investigations in the frontier of
nonlinear network theory and signal processing, computational neurobiology
and non-invasive medical diagnosis. In 1994 and 1995 he spent a year at
the Tokushima University (Tokushima, Japan) as a visiting scholar working
on cellular neural network projects related to medical image processing.
In 1997 and 1998 he conducted research in nonlinear image processing and
neuromorphic modeling of the vertebrate retina at the University of
California (Berkeley, USA). He received the PhD degree in electrical
engineering from the Budapest University of Technology and Economics in
1999. Recently, along with his co-authors, he has won the Best Paper Award
for 1998 from the Wiley Publisher (CTA). Currently his research interest
is focused on computational aspects of cellular nonlinear arrays and
includes neuromorphic modeling of biological sensing, nonlinear adaptive
techniques in signal processing and special topics in machine vision. He
is the member of the IEEE and a technical committee member of NSA-TC (IEEE
NN Society), CNNAC-TC and SMM-TC (IEEE CAS Society).
Ákos Zarándy was born in 1967. He received the M.S. degree in electrical
engineering from the Technical University of Budapest in 1992. He joined
to Analogic and Neural Computing Systems Laboratory of the Computer and
Automation Research Institute of the Hungarian Academy of Sciences in
1990. He completed his PhD studies between 1992-1997. During this time he
was dealing with Cellular Neural/nonlinear Networks (CNN) programming, CNN
chip development system building, biological modeling, color processing,
modeling of visual illusions, and mathematical morphology. He spent more
than two years at the University of California (Berkeley, USA) as visiting
scholar and as visiting post doc. Currently, his research interest is
focused on the building of ultra-high speed visual decision making
systems, and biologically motivated locally adaptive sensor matrixes. He
is member of the IEEE and technical committee member of CNNAC-TC (IEEE CAS
Society).
(Back to top of IJCNN II)
|
|
(Back to top of
FUZZ III)
FUZZ III
"Data
Mining, Modeling and Knowledge Discovery in Bioinformatics"
Nik Kasabov
Director, Knowledge Engineering and Discovery Research Institute KEDRI
Auckland Keywords: soft
computing, neuro-fuzzy systems, bioinformatics, gene expression data, DNA,
proteins, gene networks
Abstract: Bioinformatics is a new area of science that is concerned
with the information processing of biological information. The first part of
the tutorial presents a short introduction to molecular biology and
Bioinformatics [1] as well as a brief review of the principles and the
techniques of soft computing [2]. Methods of learning from data, that
includes: inductive and transductive systems, adaptive learning, data and
knowledge integration are reviewed. A comprehensive software environment
NeuCom (http://www.theneucom.com) for
data mining, modeling and discovery will be demonstrated and will be made
available to participants. In the second part of the tutorial, problems in
bioinformatics and medical decision support are defined and solved with the
use of various soft-computing methods and their combination. The problems
include: promoter recognition; gene expression data analysis for cancer and
other disease profiling; modeling gene regulatory
networks; integrating gene and clinical data; protein structure
prediction; cardio-vascular risk prognosis; renal function evaluation;
survival prognosis. There are five main phases of information processing and
problem solving in most bioinformatics systems that are covered for each of
the above problems, namely: data pre-processing and filtering; feature
evaluation and feature selection; model creation and evaluation; knowledge
extraction; adaptation to new data.
References:
[1] N.Kasabov, Evolving Connectionist Systems: Methods and Applications in
Bioinformatics, Brain Study and Intelligent Machines, Springer Verlag, 2002
(www.springer.de)
[2] N.Kasabov, Foundations of neural networks, fuzzy systems and knowledge
engineering, MIT Press, 1996
Vita: Professor Nik Kasabov, FRSNZ, FNZCS, SMIEE, is the Founding
Director and the Chief Scientist of the Knowledge Engineering and Discovery
Research Institute KEDRI, Auckland (www.kedri.info/). He holds a Chair of
Knowledge Engineering at the School of Computer and Information Sciences at
Auckland University of Technology. He holds MSc and PhD from the Technical
University of Sofia. His main research interests are in the areas of:
intelligent information systems, soft computing, neuro-computing,
bioinformatics, brain study, speech and image processing, novel methods for
data mining and knowledge discovery. He has extensive academic experience,
holding positions of Associate Professor (Technical University of Sofia,
Bulgaria), Senior Lecturer (University of Essex, UK), and Professor and
Associate Professor (University of Otago, New Zealand). He is a board member
of the Asia Pacific Neural Network Assembly (APNNA) and was its President in
1997/98. Kasabov is on the editorial boards of 7 international journals and
has been on the PC of 50 international conferences in the last 5 years. He
chaired the series of ANNES conferences (1993-2001). More information of
Prof. Kasabov can be found at the Web site:
http://www.kedri.info.
(Back to top of FUZZ III)
"INTRODUCTION TO
CLUSTERING TECHNIQUES"
Katsuhiro Honda,
Osaka Prefecture University, Japan
Prof. Francesco Masulli,
University of Pisa, Italy
and
Prof. Stefano Rovetta
University of Genova, Italy
Clustering is pervasively used in several fields such as data mining,
pattern recognition, and machine learning. Clustering helps in providing
structure to large datasets, enabling for instance indexing, searching and
categorization of document databases, or exploratory models for molecular
data such as gene expression levels in disease characterization.
In this tutorial we will outline some recently proposed techniques, which
address issues of widely used clustering algorithms.
GENERALIZATION OF POSSIBILISTIC CLUSTERING
We consider a class of clustering algorithms from a unified perspective,
focused on the memberships rather than on the cost function. Then the
concept of graded possibility is introduced.
This is a partially possibilistic version of the fuzzy clustering model,
as compared to Krishnapuram and Keller’s possibilistic clustering. We
outline a basic graded possibilistic clustering algorithm and highlight the
different properties attainable by means of experimental demonstrations.
This clustering model, which features as two special cases the standard
Fuzzy c-Means and the Possibilistic Clustering algorithms, can be tuned to
represent different data models, from partitional to multimodal and
in-between. Moreover, the problem of outlier rejection can be tackled by a
suitable tuning of the algorithm.
GENERALIZATION OF LINEAR FUZZY CLUSTERING
Linear fuzzy clustering is one of attractive extensions of FCM-type
clustering method and partitions a data set into several linear clusters
using linear varieties as the prototypes of clusters. Because the vectors
spanning the prototypical linear varieties are the eigenvectors of fuzzy
scatter matrices, the linear clustering algorithms are often identified with
a simultaneous approach to fuzzy clustering and local principal component
analysis (Local PCA).
In this tutorial, several approaches to the generalization of linear fuzzy
clustering are shown. Considering the similarity with PCA, the objective
function for linear fuzzy clustering is defined by using the lower rank
approximation of data matrix and minimized with the optimization techniques
used in PCA.
The modification enables to handle missing values or component-wise
outliers that are difficult to deal with in the conventional algorithms. The
above observation also implies a close relation between linear fuzzy
clustering and Local PCA based on probabilistic mixture models.
We shall show a new paradigm of local PCA based on fuzzy modeling
approaches.
A DIVISIVE CLUSTERING ALGORITHM BASED ON RANK
An alternative to clustering by partitioning the data space is clustering
by a hierarchy of partitions, or Hierarchical Agglomerative Clustering. This
is the main tool used in some applications (for instance DNA microarray data
analysis is routinely performed by this technique), since it produces trees,
which are easy to analyze visually, but is not immune from problems.
Being an agglomerative method, one that considers just a pair of objects
at each iteration, it is sensitive to small perturbations of data (which
produce different hierarchical trees) and does not provide hints on the
reasonable depth of the tree (under which further subdivisions are mainly
due to noise).
We describe a different clustering technique, which is divisive and based
on clustering the ranks of the distance matrix. This technique gets rid of
the disadvantages above.
Vitas:
KATSUHIRO HONDA
Katsuhiro Honda is currently a Research Associate, Department of
Industrial Engineering, Graduate School of Engineering, Osaka Prefecture
University, Osaka, Japan.
He received the Bachelor of Engineering and the Master of Engineering in
Industrial Engineering from Osaka Prefecture University, in 1997 and 1999,
respectively.
His research interests include hybrid techniques of fuzzy clustering and
multivariate analysis, data mining with fuzzy data analysis and neural
networks.
He is the author or co-author of several scientific papers, published by
journals and conference proceedings.
In 2002, he received the Best Paper Award from Japan Society for Fuzzy
Theory and systems discussing a thesis on the close relation between fuzzy
clustering and probabilistic mixture models.
FRANCESCO MASULLI
Francesco Masulli is an Associate Professor of Computer Science with the
University of Pisa (Italy). He received the Laurea degree in Physics from
the University of Genova in 1976. After the military service, he was a
researcher with the Italian National Institute for Nuclear Physics
(1978-1979), and with the Ansaldo Automazione Co. (1979-1983), and an
Assistant Professor with the University of Genova (1983-2001). He was also
in leave as a visiting scientist at the University of Nijmegen (Holland) in
1983, and at the International Computer Science Institute in Berkeley,
California in 1991, 1993, and 1994.
He authored or co-authored more than 100 scientific papers on Machine
Learning, Neural Networks, Fuzzy Systems and Ensemble Methods and co-edited
three books and two special issues of scientific journals on those subjects.
He serves as an Associate Editor the international journal "Intelligent
Automation and Soft Computing". His previous duties include the chairing of
the Conference of the International Graphomomics Society (IGS) in 1997, of
the Symposium on Soft Computing (SOCO), in 1999, and the co-chairing of the
2002 Course of the International School on Neural Networks "E.R.Caianiello"
on "Ensemble Methods in Learning Machines”, of the Workshop on Fuzzy Logic
(WILF) in 1995, 1999, 2001, and 2003, and of a special session on
Bioinformatics at IJCNN 2003. He had a tutorial on “Learning with multiple
machines: ECOC models vs. Bayesian Framework” at IJCNN 2003.
He is member of the Italian Chapter of the IEEE-Neural Network Society, an
affiliate member of the Berkeley Initiative on Soft Computing (BISC), and a
Board Member of the Italian Neural Network Society (SIREN) and of the SIG
Italy of the International Neural Network Society (INNS).
STEFANO ROVETTA
Stefano Rovetta is an Assistant Professor at the Department of Computer
and Information Sciences, University of Genoa, Italy.
He has a Laurea degree and a PhD in Electronic Engineering (from the
University of Genoa) and has worked for several years in the field of neural
networks.
His recent research interests cover applications of clustering, neural
networks, and other machine learning techniques to multimedia, biological,
and medical problems.
(Back to top of FUZZ III)
"Fuzzy Sets for Words: Why Type-2 Fuzzy Sets Should be Used and How They
Can be Used"
Jerry M. Mendel
University of Southern California
Integrated Media Systems Center
Electrical Engineering Department
Los Angeles, CA
Keywords: Computing with words, type-2 fuzzy sets, interval type-2
fuzzy sets, fuzzy logic systems
Abstract
This tutorial begins with an explanation of the provocative statement
that using type-1 fuzzy sets to model words is scientifically incorrect. It
then proposes a modified fuzzy set for a word, one that immediately includes
the intra- and inter-uncertainties that a person and a group of people have
about a word. This modified fuzzy set is an interval type-2 fuzzy set. The
rest of the tutorial then describes interval type-2 fuzzy sets and their use
in rule-based type-2 fuzzy logic systems. It includes mathematical
descriptions of and computations with type-2 fuzzy sets. Applications will
be drawn from pattern classification, time-series forecasting, knowledge
mining using IF-THEN surveys, and computing with words.
This tutorial will provide attendees with: (1) new tools that they can
immediately use in a broad collection of applications—those that have
uncertainties associated with them; (2) a good understanding of what
“computing with words” is all about; and, (3) ideas for new research
directions within the field of fuzzy sets and systems.
Vita: Jerry M. Mendel received the Ph.D. degree in electrical
engineering from the Polytechnic Institute of Brooklyn, Brooklyn, NY.
Currently he is Professor of Electrical Engineering and Associate Director
for Education, Outreach and Student Affairs of the Integrated Media Systems
Center at the University of Southern California in Los Angeles, where he has
been since 1974. He has published over 440 technical papers and is author
and/or editor of eight books, including Uncertain Rule-based Fuzzy Logic
Systems: Introduction and New Directions (Prentice-Hall, 2001). His present
research interests include: type-2 fuzzy logic systems and their
applications to a wide range of problems, including target classification
and computing with words, and spatio-temporal fusion of decisions. He is a
Life Fellow of the IEEE and a Distinguished Member of the IEEE Control
Systems Society. He was President of the IEEE Control Systems Society in
1986, and is presently Chairman of the Fuzzy Technical Committee and an
elected member of the Administrative Committee of the IEEE Neural Networks
Society. Among his awards are the 1983 Best Transactions Paper Award of the
IEEE Geoscience and Remote Sensing Society, the 1992 Signal Processing
Society Paper Award, the 2002 Transactions on Fuzzy Systems Outstanding
Paper Award, a 1984 IEEE Centennial Medal, and an IEEE Third Millennium
Medal. During his career he has taught more than 30 short courses.
(Back to top of FUZZ III)
|