Newsgroups: comp.robotics
Path: brunix!cat.cis.Brown.EDU!noc.near.net!paperboy.wellfleet.com!news-feed-1.peachnet.edu!gatech!howland.reston.ans.net!EU.net!sun4nl!news.nic.surfnet.nl!utciva.civ.utwente.nl!infnews.cs.utwente.nl!vet
From: vet@cs.utwente.nl (Paul van der Vet)
Subject: ECAI'94 Tutorial Programme
Message-ID: <Cs3Huv.69r@cs.utwente.nl>
Sender: usenet@cs.utwente.nl
Nntp-Posting-Host: ethanol.cs.utwente.nl
Organization: Twente University, Dept. of Computer Science
Date: Tue, 28 Jun 1994 06:54:31 GMT
Lines: 606




                                 ECAI 94


           11TH EUROPEAN CONFERENCE ON ARTIFICIAL INTELLIGENCE

                            AUGUST 8 - 12, 1994

                            TUTORIAL PROGRAMME


         AMSTERDAM RAI, INTERNATIONAL EXHIBITION AND CONGRESS CENTRE
                         AMSTERDAM, THE NETHERLANDS

Organized by the European Coordinating Committee for Artificial
Intelligence (ECCAI)

In cooperation with AAAI and IJCAI

Hosted by the Dutch Association for Artificial Intelligence (NVKI)

For information please contact:
Erasmus Forum
P.O. Box 1738
3000 DR Rotterdam
The Netherlands
Tel: +31 10 4082302
Fax: +31 10 4530784
E-mail: M.M.deLeeuw@apv.oos.eur.nl


TUTORIAL PROGRAMME

A full tutorial programme will take place on August 8 and 9,
1994. Thirteen lectures will be given by experienced instructors.
Extended tutorial information can be obtained by anonymous FTP from
swi.psy.uva.nl, directory pub/ecai94.

Tutorial Chairperson:
 Dr. Frank van Harmelen
 SWI
 University of Amsterdam
 Roetersstraat 15
 1081 WB Amsterdam
 The Netherlands
 Phone:  +31 20 525 6791 or +31 20 525 6789
 Fax:    +31 20 525 6896
 E-mail: ecai94-tutorials@swi.psy.uva.nl


SCHEDULE

Monday August 8, 09.00 - 13.00 hrs

T1  Models of uncertainty and graduality in AI
    Didier Dubois and Philippe Smets 

T2  The knowledge medium: the use of formal knowledge representation
    for institutional memory and communication
    Thomas Gruber and Luc Steels

T3  Intelligent multimedia interfaces
    Mark Maybury and Yigal Arens 


Monday August 8, 14.00 - 18.00 hrs

T4  Reasoning with cases: theory and practice
    Klaus-Dieter Althoff, Michel Manago and Stefan Wess 

T5  The art and the science of modelling: crucial issues in building
    second generation knowledge-based systems
    Peter Struss and Bert Bredeweg

T6  Validation of knowledge-based systems
    Pedro Meseguer and Alun Preece 


Tuesday August 9, 09.00 - 13.00 hrs

T7  Managing machine-learning application development and
    organisational implementation 
    Yves Kodratoff and Vassilis Moustakis 

T8  Knowledge-based production management
    Norman M. Sadeh and Stephen F. Smith 

T9  Multi-agent systems and distributed AI
    Les Gasser and Jeffrey Rosenschein 

T13a Artificial life and autonomous robots (theory)
     Luc Steels and David McFarland


Tuesday August 9, 14.00 - 18.00 hrs

T10  Temporal reasoning in AI
     Han Reichgelt and Lluis Vila 

T11  Rules in databases and knowledge bases
     Ulrike Griefahn and Rainer Manthey 

T12  Principles and practice of knowledge aquisition
     Angel R. Puerta and Henrik Eriksson 

T13b Artificial life and autonomous robots (practise)
     Luc Steels and David McFarland



T1   Models of uncertainty and graduality in AI

Didier Dubois (National Center for Scientific Research, France) and
Philippe Smets (Institut de Recherches Interdisciplinaires et de
Developpements en Intelligence Artificielle, France)

Researchers in automated reasoning, in database management systems and
in knowledge-based systems have felt the need for techniques that cope
with imperfect information. This is especially true when issues such
as inconsistency handling, numerical/symbolic interface, and belief
revision are addressed. Classical logic and the Bayesian view of
probability are not always enough to deal with those issues even if
they suggest useful guidelines.

This tutorial will provide an introduction to non classical models of
uncertainty and vagueness that have been developed in the last 20
years, very often in connection with Artificial Intelligence. These
models include numerical approaches such as fuzzy sets and possibility
theory, belief functions, and more logical-oriented developments such
as nonmonotonic reasoning. The tutorial is made of four lectures
respectively devoted to a survey of various forms of imperfect
information, an introduction to fuzzy sets, fuzzy logic and
possibility theory, a comparative introduction to several belief
functions theories, and the connection between uncertainty models and
nonmonotonic inference. The tutorial will survey the various
approaches in a knowledge-based systems perspective, hoping that it
might lead potential users to a better understanding of why these
models differ, how they can be related and when to use what.

Prerequisite Knowledge: The tutorial will assume that the audience has
some knowledge of probability theory and propositional calculus.


T2 The knowledge medium
   (The use of formal knowledge representation for institutional
   memory and communication)

Thomas Gruber (Stanford University) and Luc Steels (Free University Brussels)

Computer technology has begun to redefine how human knowledge is
communicated and used in organisations. Networked multimedia systems,
which can store and display information in a variety of modalities,
are used to support the communication of virtual teams across space
and time barriers.  Knowledge systems, which do limited reasoning on
symbolic representations of knowledge, are used to deliver specialised
or complex knowledge in an operational form where it is needed. The
knowledge medium is the convergence of these two trends, in which
machine-interpretable representations are part of the medium by which
we communicate and transmit our knowledge.

This tutorial will explore the role of knowledge representation in a
medium for communication and institutional memory. Applications will
be described in which knowledge is communicated and shared in forms
understood by both humans and software agents. Examples include
interactive, model-generated documentation of designed artifacts;
content-based routing of information among collaborating agents; and
the synthesis of knowledge-based software from libraries of reusable
components. Fundamental research issues to be discussed include the
design and use of shared representations, and knowledge level
specifications of tasks, agent capabilities, and information needs.

Prerequisite Knowledge: This tutorial is intended for researchers and
practitioners interested in new applications of AI to problems of
enterprise integration, computer-supported communication, software
reuse, and software interoperability.  Only a basic familiarity with
knowledge representation and software engineering will be assumed.


T3   Intelligent multimedia interfaces

Mark Maybury (MITRE Corporation, Badford, MA, USA) and Yigal Arens
(University of Southern California)

The purpose of this tutorial is to introduce the emerging literature
and set of techniques for building multimedia and multimodal
interfaces, i.e., those interfaces that interpret and generate
multiple media, e.g., spoken and written natural language, graphics,
non-speech audio, maps, animation). This tutorial will last a half day
and will be primarily a lecture, with time allotted to clarify issues
and respond to questions from the audience.

This tutorial will begin by motivating the value of a system that is
able to communicate using multiple media and modalities using examples
found in human-human communication. We define terms and note
terminological problems found in the literature, then describe an
architecture for integrated multimedia parsing and generation, which
will serve as a reference model for the remainder of the tutorial. We
next distinguish the specific kinds of knowledge utilised by these
systems and then describe the theory, illustrated with implemented
application examples, of multimedia parsing and generation. Finally,
we discuss systems that have integrated both parsing and
generation. The tutorial concludes by outlining key areas for further
research and expected future directions in the field. Throughout the
tutorial, specific descriptions of prototype systems (e.g., from the
MIT Media Lab, USC/ISI, Columbia University, the German Center for
Research in AI, IRST) will be used to illustrate the components of a
more general model of multimodal and multimedia communication.

Prerequisite Knowledge: This tutorial is relevant to those researchers
and practitioners interested in investigating, designing, and
implementing intelligent interfaces that exploit multiple media and
modalities to facilitate human- computer interaction. There is no
prerequisite knowledge required, although general knowledge of user
interfaces and artificial intelligence will enhance the value of this
course for participants.


T4   Reasoning with cases: theory and practice

Klaus-Dieter Althoff (University of Kaiserslautern), Michel Manago
(AcknoSoft, France), Stefan Wess (University of Kaiserslautern)

The objective of the tutorial is to present two technologies for
reasoning with cases: induction and case-based reasoning. Induction is
a form of machine learning that is used to automatically extract
general knowledge (for instance in the form of a decision tree or a
set of rules) from a database of cases. Case-based reasoning is a
problem solving method that makes direct use of past experiences
(cases) rather than a corpus of general knowledge such as rules.

In this tutorial, we will show how reasoning with cases helps solve a
new category of applications and how it also offers an alternative to
classical rule-based reasoning. We will introduce, compare and
contrast the two technologies, expose the history and areas of current
research, present the architecture of a case-based reasoning system
and describe some basic algorithms. We will show how cases can be
indexed for efficient retrieval, how the similarity between new and
past cases is assessed, how cases are represented (feature-value
vectors, object representations), how to use additional background
domain knowledge, and we will compare the technologies with other
forms of automated reasoning.

Induction and case-based reasoning are now mature technologies that
have reached the market. Strategic custom applications in various
domains have been delivered (and are being used) and "off the shelf"
products are available. We will review tools developed by commercial
and non-commercial organisations, identify the market for these and
show some real applications in technical maintenance and diagnosis.

Prerequisite Knowledge: This tutorial aims at presenting a survey of
the technologies and delineating their areas of application. The
intended audience is composed of the managing and technical staff of
computer divisions interested in technologies for reasoning with
cases. Knowledge engineers interested in up-to-date methodologies for
developing applications and users with a specific potential
application in mind will also appreciate the tutorial. There are no
prerequisites.


T5   The art and the science of modelling:
     (crucial issues in building second generation knowledge-based systems)

Peter Struss (Technical University of Munich) and Bert Bredeweg
(University of Amsterdam)

Reasoning about the physical world has always been a key problem in
AI. It is in the core of common sense reasoning, and it is central to
many automated problems solvers that are intended to deal with
industrial applications. Recently, model-based systems have become a
focal point of both theoretical work and efforts to build powerful
systems, and the field is now mature for significant applications.
Designing an adequate model for the domain and task at hand is the key
problem and step. The quality of the model crucially effects the
competence and robustness of the problem solving system, the
generality and reusability of the knowledge base (and, hence,
development costs) the interaction with the system and its performance
(e.g.  real-time behaviour). The research efforts of the last fifteen
years or so result in a vast set of powerful theories, useful
techniques, and sophisticated systems which is hard to overlook for
the practioner and the newcomer to the field, and, still, much of
successful modeling appears to be more like an art rather than an
engineering task. The tutorial provides a critical overview of the
field. It discusses requirements and objectives in modeling for
knowledge-based systems, the existing formal theories and tools as
well as their limitations and open problems. We will organise it along
a number of key problems, questions, and requirements raised by real
domains and applications, and analyze how different theories and
techniques address these issues. Examples deal with design and
configuration, failure mode analysis, simulation, analysis, testing,
sensor placement, diagnosis and repair, supervision, explanation and
tutoring and treat physical systems, biological and ecological
systems, and enterprises.

Prerequisite knowledge: The tutorial is intended for developers of
industrial applications as well as novices in the field and
researchers from other AI areas. The former should receive help for
finding solutions to their problems, while the latter may find what is
worth while working on. No extensive knowledge in AI is required for
attending. Some basic knowledge in mathematics, logic, knowledge
representation, and reasoning is helpful, but not mandatory.


T6   Validation of knowledge-based systems

Pedro Meseguer (Technical University of Catalonia, Barcelona) and Alun
Preece (University of Savoie, France)

This tutorial will provide participants with a firm understanding of
the current state-of-the-art in techniques and tools for validating
knowledge-based systems (KBS). Highlights of the presentation will
include a detailed examination of methods for performing rigorous
verification, testing and evaluation of these systems. Our intention
is to give attendees a solid introduction to the theoretical
foundation of KBS validation methods, together with a clear
understanding of how the methods can be applied in practice. Our
analysis of KBS verification tools and techniques will cover not only
classical rule-based systems, but also systems with uncertainty,
explicit control knowledge, frames, and procedural components. We will
describe the underlying algorithms, and will draw upon practical
experience to consider issues arising in the use of verification
tools---for example, how to interpret their output. A thorough survey
of testing techniques for KBS will include both an assessment of the
applicability of testing approaches from software engineering, and an
examination of special problems in testing KBS. We will also consider
broader issues in KBS evaluation, concerned with ensuring that a
delivered system will be used productively. All of these techniques
will be illustrated using case studies from KBS practice. Finally, we
place the validation activities in context by relating them to other
activities in the KBS life-cycle, and also by relating them to other
important topics in artificial intelligence. Participants will leave
this tutorial with a set of recommendations for carrying out rigorous
and effective validation to ensure the quality of their KBS.

Prerequisite Knowledge: Participants in this tutorial should be
familiar with knowledge-based systems, at least to the level of an
introductory textbook. The tutorial is aimed at participants with an
interest in quality assurance of KBS, especially systems builders or
managers currently involved--or anticipating involvement--in
developing KBS.


T7 Managing machine-learning application development and
   organisational implementation

Yves Kodratoff (University of Paris-Sud) and Vassilis Moustakis
(Technical University of Crete, Greece)

Applying machine learning (ML) techniques to Industry is made at once
difficult by the large amount of available techniques and programs
that are able to perform induction or support learning of some
kind. It also represents a challenge both in terms of the application
and of ML itself in that a great proportion of the user community is
torn between the hopes and promises brought about by innovations in ML
science and technology on the one hand and their need to understand,
and, ultimately use these innovations to support knowledge based
system (KBS) tasks on the other.  The tutorial will cover all phases
underlying (ML) application development. Special emphasis will be
placed upon lessons learned from existing industrial `real world' ML
applications. From a technical point of view the tutorial will address
the issue of coupling application specifics with ML systems. It will
suggest a systematic framework, sufficient for supporting ML
application management and ML system selection according to
application requirements. It will also emphasise the numerous problems
met when Knowledge Acquisition has to be coupled with ML: most
existing industrialised ML applications had to perform the acquisition
of the knowledge necessary to have the ML system running.

The tutorial will adopt an application bias in reviewing the
conditions under which a given ML technique should be applied. Results
are demonstrated by way of a series of real world ML application case
studies.  The tutorial should be useful to both ML researchers and ML
practitioners. ML practitioners will get an overview of ML system
capabilities and potential in addition to having a chance to learn
about real world applications. ML researchers may find this tutorial
useful in understanding reality of applications and identify gaps
pointing to the need for further system development or enhancement.

Prerequisite Knowledge: Participants who want to attend this tutorial
should possess a minimum of AI skills. Acquaintance with machine
learning is desirable although not necessary.


T8 Knowledge-based production management

Norman M. Sadeh and Stephen F. Smith (Carnegie Mellon's Robotics
Institute, Pittsburgh)

This tutorial will introduce participants to the concepts, techniques,
and methodologies that have emerged from work in knowledge-based
production management. We will first consider the shortcomings of
traditional approaches to production management (e.g., MRP/MRP II) and
identify opportunities provided by knowledge-based technologies both
in overcoming these limitations, and in contributing to effective
implementation of modern manufacturing philosophies (e.g. Just in
Time).  We will then review in more detail the essential concepts and
techniques underlying dominant approaches to knowledge-based
production management. We will cover object-centered modeling
frameworks, simulation and rule-based techniques, temporal constraint
management, blackboard and multi-perspective techniques, constrained
heuristic search, uncertainty management, iterative improvement
techniques, distributed production management, and intelligent
interactive scheduling frameworks. In each case, we will characterise
strengths and weaknesses from the standpoint of different production
management requirements, and indicate the results that work under each
approach has produced to date.  Finally, we will examine a few
successful applications, and assess the current state of theory and
practice.

Over the past decade, a large (and continually increasing) number of
efforts (both research and development) have sought to investigate and
exploit the use of Artificial Intelligence (AI) concepts and
techniques in production management applications. In some cases,
AI-based concepts have provided frameworks for making traditional
Operations Research (OR) techniques more accessible and usable in
practical production management settings. In other cases, novel
concepts and techniques have been developed that offer new
opportunities for more cost-effective factory
performance. Knowledge-based scheduling and planning techniques are
having an increasing operational impact in complex production
management applications.

Prerequisite Knowledge: This tutorial is aimed at both practitioners
and researchers who are interested in applying knowledge-based
techniques to practical production management problems. It will also
be useful to technology managers who want to keep abreast of the
current state of the art in knowledge-based production management. The
tutorial assumes knowledge of AI at the level of an introductory
course as well as some familiarity with basic production management
concepts.


T9   Multi-agent systems and distributed AI

Les Gasser (University of Southern California) and Jeffrey
S. Rosenschein (Hebrew University Jerusalem)

Multi-Agent Systems and Distributed AI (MAS/DAI) are concerned with
how to coordinate behaviour among a collection of semi-autonomous
problem-solving agents, so that they can act together to solve joint
problems, or make individually or globally reasonable decisions
despite uncertainty and conflict.  MAS/DAI systems are a research
reality, and are rapidly becoming practical partners in critical tasks
such as telecommunications control and management, power distribution,
product development, manufacturing, robotics, enterprise
integration/coordination, and organization design.

This tutorial will provide a thorough survey of problems, theory,
techniques and applications in contemporary Multi-Agent Systems and
Distributed AI. We will develop a comprehensive picture of current
knowledge and contemporary currents in MAS/DAI, in preparation for
building MAS/DAI systems or as background for doing advanced research
on outstanding MAS/DAI problems. The tutorial is designed for people
who are professionally interested in building MAS/DAI systems, for AI
researchers interested in learning about a range of MAS/DAI
approaches, and for technology planners and managers who need to know
about leading-edge technologies.

Prerequisite Knowledge:
The tutorial presumes knowledge of AI at the level of an introductory
AI course.


T10   Temporal reasoning in AI

Han Reichgelt (University of the West Indies in Mona, Jamaica) and
Lluis Vila (Institute for Research in AI of Blanes, Spain)

The notion of time is ubiquitous in any activity that requires
intelligence. A whole range of intelligent tasks require reasoning
about time like Diagnosis, Explanation, Planning, Process supervision,
Natural language understanding. It follows that the representation of
time and reasoning about time is of crucial importance for Artificial
Intelligence systems. This tutorial is intended to demonstrate it, to
give a clear picture of the different issues involved in a temporal
reasoning system, and to provide a progressively detailed analysis of
each of these different issues discussing the advantages and
shortcomings of the different approaches in the literature. The
tutorial will be comprised of the following sessions:
  Introduction,
  What is temporal reasoning?,
  Why it is so important?,
  How it can be used in practice?,
  Ontologies of time,
  How should time be conceptualised, e.g. as points or as intervals?,
  What are the advantages of each conceptualisation?,
  Temporal logics,
  Method of temporal arguments,
  Modal temporal logic,
  Reified temporal logic,
  Critical comparison of the three 
  Algorithms for temporal reasoning,
  Change, causality and non-monotonicity.    

Prerequisite Knowledge: The tutorial will assume some basic
understanding of first-order predicate calculus. However, it will
assume no in-depth knowledge of the field of temporal reasoning. It is
our intention to present both introductory and in-depth material. The
tutorial will therefore be suitable both to complete novices in the
field and those with some background in the area.


T11   Rules in databases and knowledge bases

Ulrike Griefahn and Rainer Manthey (University of Bonn)

Rules have been a well-known concept in artificial intelligence since
long, investigated in connection with expert systems, knowledge bases,
or logic programming. Various instances of the rule concept were later
adopted by the database community in view of extending AI techniques
towards the handling of large amounts of data. Today there are two
major directions of activity in database research which are concerned
with the introduction of rules into database systems: active and
deductive databases.  Meanwhile, a significant amount of
database-specific techniques have been developed. These approaches are
relevant for research in AI, too, as many of today's expert system
applications are related to large quantities of data. But even in case
of comparatively small amounts of data, not necessarily requiring
database technology, some of the techniques developed in the DB
community seem to be interesting alternatives to "classical" AI
solutions.  The tutorial aims at providing a compact, up-to-date
overview of the state-of-the-art in active and deductive databases. In
addition, first results towards an integration of both kinds of rules
within a common framework are presented, focusing on the
implementation of deductive inference by means of active rules.

Prerequisite Knowledge: The presenters try to keep the tutorial
largely self-contained. However, basic knowledge about databases
(primarily relational databases), logic programming, and expert
systems are helpful.


T12   Principles and practice of knowledge acquisition

Angel R. Puerta (Stanford University) and Henrik Eriksson (Linkoeping
University)

Knowledge acquisition draws from many research areas. Due to the
complexity of the fields involved, it is often more productive to
examine knowledge acquisition from an empirical point of view, than to
do so from a purely theoretical one. This tutorial will develop a
comprehensive view of the most important principles and practical
issues in knowledge acquisition. We will present the theoretical
foundations of knowledge acquisition to establish a framework in which
the attendee can understand and analyze how the theories are put to
practice. We will concentrate on illustrating problems in using
computer-based knowledge- acquisition tools, covering examples from
early expert systems to the new generation of knowledge- based systems
based on reusable knowledge components. Throughout the tutorial, we
will emphasise the particular issues that affect the design and
development of knowledge-acquisition tools. The attendee will learn
what principles and design tradeoffs are involved in the construction
of knowledge-acquisition tools, and what are the research issues in
knowledge acquisition.

Prerequisite Knowledge: This tutorial is suited for anyone who has an
introductory background in artificial intelligence. The course will be
especially helpful to knowledge engineers involved in the development
of knowledge-based systems, to research scientists who work with
knowledge bases, and to anyone desiring an overview of the advances in
knowledge acquisition.


T13a and T13b   Artificial life and autonomous robots 

Luc Steels and assistants (Free University Brussels) and
David McFarland (University of Oxford)

This tutorial is a unique opportunity for AI researchers to get an
overview of the newly developing paradigm of behaviour-oriented AI and
to understand the approach in sufficient technical detail. This
tutorial consists of two parts: the first part is given in the morning
and is particularly important for the necessary background and
motivation (Steels and McFarland). The second part is given in the
afternoon and will give hands-on experience, and potentially will give
researchers the chance to continue working in this area (Steels and
assistants). Artificial Life studies the phenomenon of life the same
way AI studies intelligence: by building artificial systems that show
the same capabilities as living systems. This tutorial gives an
overview of research in Alife, the behaviour-oriented approach to AI,
and the biological significance of this research. The first tutorial
is theoretical. An overview of the literature will be given and
important general trends discussed. The second tutorial focuses on
technical issues. An autonomous robot lab will be set up in which
participants have the opportunity to build their own robots out of
components made available in the lab. The secon tutorial will be
restricted to a smaller audience.

Prerequisite Knowledge: The tutorial is intended for researchers or
developers in any area of AI that want to learn about the new exciting
developments in behaviour-oriented AI research and its interaction
with Artificial Life. There are no prerequisites beyond general
knowledge about AI. The second tutorial requires general programming
experience but no prior experience in robot building.

-- 
---------------------------------------------------------------------
Paul van der Vet                   Phone +31 53 89 36 94 / 36 90
Knowledge-Based Systems Group      Fax   +31 53 33 96 05
Dept. of Computer Science          Email vet@cs.utwente.nl
University of Twente
P.O. Box 217
7500 AE  Enschede
The Netherlands
---------------------------------------------------------------------
