The Sixth IEEE International Conference on Fuzzy Systems
July 1-5, 1997
Tutorials
Speaker:Petr Hájek, ICS, Prague, Czech Republic
Description:
The formal system of fuzzy predicate logic (logic with quantifiers) will be presented and
problems of its axiomatization will be discussed. We shall, among other things present
Pavelka-Lukasiewicz predicate logic as a calculus for proving partially true conclusions
from partially true assumptions.
This calculus will be used to analyze various items from the agenda of fuzzy logic in
terms of logical deduction, as:
- a fuzzy logic of probability, separating explicitly probability from fuzzines and building a bridge between them,
- an analysis of Zadeh's fuzzy modus ponens showing that if put properly it can be
presented as logical deduction preserving partial truth,
- an analysis of Mamdani style fuzzy controller showing that it does not concern
implication but does concern deduction.
Speaker: Takeshi Yamakawa, Kyushu Institute of Technology, Iizuka, Japan
Description:
This Tutorial Lecture presents the following contents.
- Introduction
- What is chaos?
- How is chaos produced?
- Period bifurcation
- Measure of chaos
- Chaos chip
- Application to separation of overlapped patterns by a chaos network
- Conclusions
Speaker: Daniele Mundici, Dept. of Computer Science, University of Milan, Italy
Description:
Nonboolean partitions (henceforth, partitions) are an interesting object of investigation in at least two respects:
- 1. In the ususal if-the-else approach to control, when we logically express the situation:
if H1 then A1, if H2 then A2..., if Hn then An, hypotheses H1,...,Hn governing
actions A1,...,An are propositions in an infinite-valued logic, forming a partition.
- 2. Partitions are a one-dimensional recapitulation of more complicated, two-
dimensional, objects---having to do with approximating equality.
In order to capture the fact that the components of any partition must be irredundant, we
use the machinery of the infinite-valued calculus of Lukasiewicz, and its algebraic
counterpart given by C.C.Chang's MV-algebras. Using the fact that in every MV-algebra
one has a genuine notion of sum---whence a natural notion of irredundancy, namely
linear independence---we give a rigorous and simple definition of partition, and then
establish for this notion several fundamental properties, generalizing the basic properties
of Boolean partitions, such as for instance:
- Joint refinability: any two partitions have a common refinement
- Richness: partitions have maximal expressive power
We finally show how our notion of irredundancy results in computational efficiency.
Speaker: Piero P. Bonissone, General Electric CRD, Schenectady, NY 12309, USA
Description:
The purpose of this tutorial is to introduce the concept of Fuzzy Logic based controllers
or Fuzzy Controllers (FCs), one of the most promising emerging technology in the field
of Engineering.
This tutorial is subdivided into three major parts: 1) FLC technology, 2) FLC
applications, and 3) FLC within the scope of Soft Computing. In the first part we will
illustrate the development process common to Fuzzy Systems (both fuzzy rule based
systems and fuzzy controllers). This will be followed by a detailed description of fuzzy
controllers technology and its components: FLC interpreters, compilers, and run-time
engines.
In the second part, we will discuss some illustrative FLC applications. We will compare
these applications in a cost/complexity framework, and examine the driving factors that
led to the use of FLCs in each application. We will emphasize the role of fuzzy logic in
developing supervisory controllers and in maintaining explicit the tradeoff criteria used to
manage multiple control strategies.
Finally in the third part of this report we will show how FLC and in general Fuzzy
Systems are components of a broader paradigm called Soft Computing (SC). SC is a
new discipline that combines emerging problem-solving technologies such as Fuzzy
Logic (FL), Probabilistic Reasoning (PR), Neural Networks (NNs), and Genetic
Algorithms (GAs). Each of these technologies provide us with complementary reasoning
and searching methods to solve complex, real-world problems.
Within this broader context, we will analyze and illustrate some of the most useful
combinations of SC components, such as the use of FL to control GAs and NNs
parameters; the application of GAs to evolve NNs (topologies or weights) or to tune FL
controllers; and the implementation of FL controllers as NNs tuned by backpropagation-
type algorithms.
Speaker: Hamid Berenji, NASA Ames Research Center, USA
Description:
Fuzzy Systems technology is reaching a nearly maturing stage: many
successful applications exist, its acceptance is on the rise, and many
realize its advantages and its limitations. In this tutorial, we will
discuss the next wave of futuristic fuzzy logic based systems in terms of
intelligent systems that can learn to tune themselves, learn to
perform new tasks, learn to evolve into more powerful systems, and can
demonstrate more the qualities of truly intelligent systems.
In the first part of this tutorial, we will discuss reinforcement
learning and its current role in development of powerful learning
systems. A number of techniques such as Temporal Difference (TD
Algorithms), Prioritized Sweeping, and Q-Learning will be discussed in
details.
In the second part of this tutorial, we show the
applications of reinforcement learning techniques to fuzzy systems
in order to make them learn from experience. Algorithms such as Fuzzy
Reinforcement Learning (FRL) and Fuzzy Q-Learning (FQ-Learning) in addition to
architectures such as GARIC and its very recent extensions GARIC-RB
(for Data Mining) and GARIC-Q (for multi-agent learning) will be fully
discussed.
In the third section, this tutorial will discuss other major
techniques such as Genetic Algorithms, Supervised
Learning, and other Neural Network techniques for learning in fuzzy
systems. We will compare these techniques in terms of their
strength/limitations and potentials for applications.
Last section of this tutorial will discuss a number of current
applications including examples from Space Systems (e.g., Space Shuttle
Orbital Operations), aircraft control, and Hybrid Electric Vehicles.
Speakers: Patrick Bosc, IRISA/ENSSAT, Lannion, France and Gabriella Pasi,, CNR/ITIM, Italy.
Description:
PART A: INTRODUCTION
- Database principles (relational and OO models)
- Information retrieval models
- Basic notions on fuzzy sets and possibility theory
- Fuzziness in databases (imprecise/uncertain data, soft properties, confidentiality, flexible querying)
- Fuzziness in Information retrieval (document representation, flexible
querying, soft retrieval mechanisms, soft associative mechanisms)
PART B: FUZZINESS IN DBMSs
- B.1 Fuzziness in relational DBMSs
- Representation of ill-known information
- Fuzzy functional dependencies
- Use of fuzzy predicates in queries
- Fuzzy pattern matching
- Regular relational databases and flexible querying (extending the relational algebra, extending an SQL-like query language, query processing)
- Prototypes
- B.2 Fuzziness in object oriented DBMSs
- Imprecision and uncertainty management in OO databases (data values, properties, hierarchical relationships similarities in OODMs)
- Some fuzzy OO data models (UFO, Tanaka, FOOD, Buckles&Petry ......)
PART C: FUZZINESS IN IRSs
- C.1 Fuzzy document representation
- Fuzzy indexing
- Fuzzy representation of structured documents
- C.2 Fuzzy extended Boolean models
- Schema of an extended Boolean model
- Fuzzy document representation and Boolean query language
- Fuzzy extensions of the Boolean query language
- numeric query weights
- linguistic query weights
- soft aggregation operators
- C.3 Fuzzy associative mechanisms
- Fuzzy thesauri
- Fuzzy clustering
- Fuzzy relevance feedback mehanisms (using genetic algorithms, neural networks .....)
- C.4 Other aspects (logic-based models, neural network based models, systems
evaluation)
Speaker:Enrique H. Ruspini, Artificial Intelligence Center, SRI International, USA
Description:
The increased ability, provided by recent technological developments, to access a wide
variety of databases has led to considerable interest on approaches to analyze large
amounts of multivariate data seeking to discover new knowledge in the form of useful
relations between the represented objects and between their characterizing variables.
Soft-computing techniques, including neural networks, evolutionary computation, and
fuzzy logic, play a significant role in many of these approaches. In combination with
other statistical, probabilistic, and approximate-reasoning methods, these techniques may
be applied to solve prediction, system control/identification, data exploration, and
classification problems.
This tutorial will present first a broad overall characterization of knowledge discovery and
data mining approaches, focusing later on soft-computing methods to rule learning,
cluster analysis, system and model identification, qualitative object description, and time-
series analysis.
Speaker: Didier Dubois, IRIT, Toulouse, France
Description:
The paradigm of constraint-directed reasoning is more and more pervasive in Artificial
Intelligence. This topic is at the crossroads between several fields of investigation such as
Operations Research, Engineering Design, Decision Analysis, and Automated
Reasoning, and any field concerned with problem-solving. However the formulation of a
problem in terms of hard constraints is not realistic. Possibility theory and fuzzy sets
offer a or prioritized constraints, including a qualitative approach to uncertainty. The aim
of this tutorial is to present this approach and to compare it to classical representation
techniques in multiple criteria decision-making under uncertainty. The techniques are
illustrated on examples in the field of scheduling.
OUTLINE
- 1. Introduction to possibility theory. Preference versus uncertainty-driven semantics.
- 2. Representation of soft constraints: fuzzy constraints and prioritized constraints, the
maximin paradigm and its justification. Limitations.
- 3. Ranking solutions under the maximin paradigm: the Discrimin and Leximin Rankings.
Links with OWA operations.
- 4. Computing optimal solutions to fuzzy constraint satisfaction problems: algorithmic
issues. Classes of easy problems: fuzzy linear programming and fuzzy PERT problems.
Scheduling under limited resources.
- 5. Fuzzy constraint satisfaction under uncertainty: Optimistic versus pessimistic solutions
Schedule
- July 1st morning (10:00 - 13:00) : T1 & T2
- July 1st afternoon (15:00 - 18:00) : T3 & T4
- July 2nd morning (10:00 - 13:00) : T5 & T6
- July 2nd afternoon (15:00 - 18:00) : T7 & T8
Each tutorial will approximately last three hours.
Tutorials announced in the same morning or afternoon will begin at the same time.
Chair
LLorenc Valverde
Universitat de les Illes Balears, Spain
dmilvg0@ps.uib.es
http://www.iiia.csic.es/conferences/fuzzieee97.html
ftp://ftp.iiia.csic.es/fuzzieee97
Updated: October, 18th 1996
IIIA
IEEE
WWW Server