cdfk@otter.hpl.hp.com (Caroline Knight) (03/11/91)
REMINDER - AISB'91 Tutorials - Cheap rate ends 15th March
Get your application for the tutorials in before Friday the 15th
March and get the cheap rate - prices all go up on the 16th!
Here is a reminder of the tutorials:
AISB'91 has 6 full-day tutorials taking place on the 16th of
April at Bodington Hall, Leeds, UK
You may select a second choice should your preferred tutorial be
full - numbers are limited to 30 maximum.
The fees for registration for the tutorials
before 15th March 1991 after 15th March 1991
Full 190 pounds 255 pounds
AISB members 140 pounds 205 pounds
full-time
students 55 pounds 105 pounds
You do not need to attend the conference to attend a tutorial.
Note that cheques and bankers drafts must be in pounds sterling
only, drawn on a British clearing bank. Please make cheques
payable to The University of Leeds. All registration forms and
payments will be acknowledged.
To obtain a registration form contact: aisb91@ai.leeds.ac.uk
Tutorial 1 Knowledge Base Coherence Checking
----------
Professor Jean-Pierre LAURENT
University of Savoie
FRANCE
Tutorial 2 Advanced Constraint Techniques
----------
Dr. Hans Werner Guesgen and Dr. Joachim Hertberg
German National Centre for Computer Science (GMD)
Sankt Augustin,
GERMANY
Tutorial 3 Functional Representation and Modeling
----------
Prof. Jon Sticklen and Dr. Dean Allemang*
Michigan State University
USA
* Universitaet Zurich, SWITZERLAND
Tutorial 4 Intelligent Pattern Recognition and Applications
----------
Prof. Patrick Wang
Northeastern University, Boston
USA
Tutorial 5 SILICON SOULS - Philosophical foundations of computing and AI
----------
Prof Aaron Sloman
Univ of Sussex
Brighton
UK
Tutorial 6 - Knowledge Acquisition
----------
Dr Nigel Shadbolt
Univ of Nottingham
UK
----------------------------------------------------------------
----------------------------------------------------------------
Here are further details of each tutorial:
Tutorial 1 Knowledge Base Coherence Checking
----------
Professor Jean-Pierre LAURENT
University of Savoie
FRANCE
Brief Description of Content
----------------------------
As well as Conventional Softwares, AI Systems need validation tools
also. Some of these tools must be specific, especially for validating
Knowledge-Based Systems, and in particular for checking the coherence
of a Knowledge Base (KB). In introduction to this tutorial we will
clarify the distinctions to be made between Validation, Verification,
Static Analysis and Testing. Among the Verification methods the issue
of checking the Coherence of a Knowledge Base is quite specific to KB
Systems. The notion of a model of KB for Coherence Checking will be
discussed and concepts of "Static Coherence" and "Dynamic Coherence"
will be introduced. The problem of Dynamic Coherence of a Knowledge
Base is the main challenge, as there is obviously a risk of
combinatorial explosion when checking for Dynamic Coherence.
First we will present methods which try to check exhaustively for
the coherence of a knowledge Base. They exhibit the specification of
all the possible initial Fact Bases that may allow infering a terminal
fact and then a human expert can attest that these initial Fact bases
are realistic or not.
Second we will present a pragmatic approach in which, instead of
trying to assert the global coherence of a KB, it is proposed to check
wether it contains incoherences. It uses a specific and heuristic
knowledge provided by the experts in the field (which we call
"Coherence Knowledge") to automatically generate a reasonable set of
"conjectures of incoherence". For each of these, a backward-chaining
proof is attempted, to assert whether the corresponding incoherence
can be deduced from the KB or not. Once no more incoherence
conjectures can be proved, the KB is "assumed to be coherent".
This approach is illustrated by the SACCO System, dealing with KBs
which contain classes and objects, and furthermore rules with
variables. This system takes into account the fact that exhaustive
coherence checking may often be impossible in practice. It also offers
the possibility to define, from the beginning of the development
phase, protocols for KB coherence checking, that will be used to
receipt the KB system or not.
Professor Jean-Pierre LAURENT was initially a member of the
University of PARIS VI Research Group. After a sabbatical period in
1979 at the Stanford Research Institute, he was Professor in 1981 at
the University of Chambery (France). He has created there the LIA
Laboratory, the main research areas of which are : Methodologies for
Developping Knowledge-Based Systems, Quality of AI Software Systems,
and Distributed Artificial Intelligence. Thirty researchers are
currently working in his laboratory. J.P. LAURENT has co-organized the
"First European Workshop on Validation, Verification and Test of KB
Systems" during ECAI 90 in Stockholm (August) and has given a tutorial
on this topic at COGNITIVA 90 in Madrid (November). He is one of the
main founders and responsible persons of the French National Program
in AI. He was a member of the "IJCAI 89 Advisory committee" and will
be the Local Chairman of the IJCAI 93 Conference that he will organize
in Chambery.
Tutorial 2 Advanced Constraint Techniques
----------
Dr. Hans Werner Guesgen and Dr. Joachim Hertberg
German National Centre for Computer Science (GMD)
Sankt Augustin,
GERMANY
Brief Description of Content
----------------------------
Constraint satisfaction describes a class of techniques that can be
applied to solve, or partially solve, networks of relations. The
venerable Waltz Filtering technique as well as Mackworth's arc and
path consistency algorithms are representatives of this class.
Classical constraint techniques, however, have been found improvable
when used in practical applications: on the one hand, new theoretical
concepts had to be developed to cope with inconsistency and with
infinite domains, cf. work of Dechter and Pearl on constraint
relaxation and the well known work of Allen on reasoning about time.
On the other hand, there has been work in new ways of implementing
constraint satisfaction exploiting the fact that a constraint network
suggests itself for parallel or massively parallel implementation, cf.
work of Cooper and Swain, and Kasif.
The tutorial will present a coherent overview over these more recent
concepts and approaches. It presents the concept of dynamic
constraints as a formalism subsuming classical constraint
satisfaction, constraint manipulation and relaxation, bearing a
relationship to reflective systems; moreover, the tutorial presents
approaches to paralllel implementations of constraint satisfaction in
general and dynamic constraints in particular.
Dr. Guesgen has published a book "CONSAT: A system for constraint
satisfaction" (Morgan Kaufmann), has presented papers on the subject
at IJCAI'87 and IJCAI'89, and has had a paper on constraint
propagation (with Dr. Hertzberg) in Artificial Intelligence. Dr.
Hertzberg is the Program Chair for the 1991 European Workshop on
Planning.
Tutorial 3 Functional Representation and Modeling
----------
Prof. Jon Sticklen and Dr. Dean Allemang*
Michigan State University
USA
* Universitaet Zurich, SWITZERLAND
Brief Description of Content
----------------------------
A growing body of AI research centres on using the known functions
of a device as indices to causal understanding of how the device
"works". The results of Functional Representation and Modeling have
typically leveraged this organization of causal understanding to
produce tractable solutions to inherently complex modeling problems.
The Functional Approach is a subfield of Model Based Reasoning (MBR).
The MBR area can be broken down into (a) techniques for obtaining a
behavioural model of a device [e.g. deKleer and Brown, Kuipers,
Bylander...] and (b) techniques for utilizing a behavioural model once
it is in hand [e.g. davis, deKleer's circuit work,...]. Techniques of
Functional Reasoning are of the second type. In this tutorial, the
fundamentals of Functional Representation and Reasoning will be
explained. Liberal use of examples through out will illustrate the
representational concepts underlying the Functional Approach.
Contacts with other MBR techniques will be made whenever appropriate.
Sufficient background will be covered to make this suitable for both
those unacquainted with the MBR field, and for more experienced
individuals who may be working now in MBR research. A general
familiarity with AI is assumed. Participants should send in with
their registration materials a one page description of a modeling
problem which they face in their domain.
Professor Sticklen is the Research Director of the AI/Knowledge Based
Systems Lab, within the Computer Science Department at Michigan State
University, and is an internationally recognised expert on model based
reasoning.
Tutorial 4 Intelligent Pattern Recognition and Applications
----------
Prof. Patrick Wang
Northeastern University, Boston
USA
Brief Description of Content
----------------------------
The following is Prof Wang's original proposal for two separate
tutorials. The actual full-day will contain elements of both.
(a) Pattern recognition and image processing have a very close
relation. They both deal with digitized images by computer. On one
hand, the image processing techniques are normally employed in pattern
recognition as "preprocessing", such as digitization, noise
elimination, image data compression, skeletonization (thinning), and
primitive selection. On the other hand, the results of pattern
recognition are helpful in solving image processing problems such as
scene analysis, image description and understanding, segmentation,
edge detection, and feature extraction. Applications of pattern
recognition and image processing include character recognition,
signature verification, target detection, medical diagnosis, remote
sensing, identification of human faces and fingerprints, speech
recognition and understanding, and machine vision and parts
inspection.
This tutorial covers the following subtopics:
* What is pattern recognition (PR)?
* The roles image processing plays in PR
* Image digitization and noise elimination
* Thinning and image data compression: sequential methods vs. parallel
vs. flexible
* Image segmentation and egde detection
* Primitive selection and feature extraction
* Imaging models
* Examples and applications.
(b) For the past decades, there is a growing interest in the study of
AI and rule-based systems. Pattern recognition plays an important
role in such systems. In fact, there is now much interaction between
expert systems and pattern analysis. It is interesting to see that
the core of pattern recognition, including "learning techniques" and
"inference" also plays an important and central role in AI. Visual
perception, scene analysis and image understanding are also essential
to robotic vision. On the other hand, the methods in AI such as
knowledge representation, semantic networks, and heuristic searching
algorithms can also be applied to improve the pattern representation
and matching techniques in many pattern recognition problems - leading
to "smart" pattern recognition. Moreover, the recognition and
understanding of sensory data like speech or images, which are major
concerns in pattern recognition, have always been considered as
important subfield of AI.
This tutorial covers the following subtopics:
* Overview of Pattern Recognition (PR) [including most recent
developments at MIT]
* Overview of Artificial Intelligence (AI) [including most recent
developments at MIT]
* The relation between PR and AI: concentrating on learning
* The concepts of Learning and Inferencing: supervised vs.
non-supervised
* The four main approaches to PR: Statistical, Syntactical, Structural
and Histogram
* Some Multi-dimensional models for PR and Object Recognition:
examples and applications of array grammars and others (including MIT)
* Degrees of recognizability, learnability, understandability and
ambiguity
* Knowledge representation and semantic networks for PR
* An example - Line-drawing PR: BM Method vs. Extended Freeman Chain
Code (EFCC) vs. Improved EFCC (IEFCC)
* Anaother example - Knowledge Pattern Representation of Chinese
Characters: hierarchical structure, induced knowledge,
syntax-semantics correlation, common patterns and logical relations
between characters, and new character principle.
Dr. Wang is a Visiting Scientist at MIT AI Lab on leave from
Northeastern University. He also teaches part-time at Harvard
University. He has edited four books, and published over sixty
technical papers in imaging technology, pattern recognition and AI.
He has given tutorials in this area at the Electronic Imaging
Conferences (EI'89, EI'90) and at the International Conference on
Pattern Recognition (ICPR'90).
Tutorial 5 SILICON SOULS - Philosophical foundations of computing and AI
----------
Prof Aaron Sloman
Univ of Sussex
Brighton
UK
Prof. Sloman's tutorial is already proving popular. If you wish to
attend please book as soon as you are able. There is a maximum of
30 attendees. All booking is being handled at leeds, contact:
aisb91@ai.leeds.ac.uk
The following is taken from the original provisional proposal. Note that
the actual tutorial will vary with the level and manner of participation.
This tutorial requires the participants to restrain themselves as well
as join in - note the prerequisites below.
Brief description
-----------------
This will not be a technical tutorial. Rather the tutor will
introduce a collection of philosophical questions about the nature
of computation, the aims of AI, connectionist and non-connectionist
approaches to AI, the relevance of computation to the study of
mind, varieties of mechanism, consciousness, and the nature of
emotions and other affective states. Considerable time will be
provided for discussion by participants.
More detailed outline
---------------------
Prerequisites
Intelligence. The ability to think about abstractions without
necessarily worrying about what their applications are. Some
familiarity with the writings of a non-empty subset of: Dennett,
Haugeland, Fodor, Pylyshyn, Dreyfus, Hofstadter, Clarke, Searle,
Penrose.
Ability to listen before talking.
Ability to take part in a discussion without hogging it or going on
and on about topics that don't interest the rest of the group.
Content
The tutorial will last a whole day and will be make up of four main
sessions separated by breaks for refreshment.
Session 1. - Breakfast till morning coffee
Approaches to the study of mind. Factual vs non-factual
questions. The space of possible designs.
Computational and other mechanisms. Virtual machines. How
computers treat _themselves_ as virtual machines. The importance
of structural variability. Limitations of vector spaces.
Causal connections between virtual mechanisms. Why the
_physical_ symbol system theory is muddled.
Is there any philosophically important difference between
GOFAI and connectionism? (No).
Session 2. Morning coffee till lunch.
What needs to be explained? Characteristics of mind:
Intentionality, flexibility, productive laziness. Why it's
silly to look for a boundary between minds and non-minds:
look for many small discontinuities, not one big one. Why most
people who discuss consciousness are fooling themselves.
What are the criteria for an adequate explanation of mind? Will
the concepts of ordinary language survive the development
of adequate theories? Disposing of some fallacious views about
the characteristics of computers.
Session 3. Lunch till tea
Interesting discontinuities in the space of possible designs.
Mechanisms required for intentionality. Semantics in computers
without AI. Structural vs causal underpinnings of meaning. The
functional requirements for meaningful use of symbols. What
kinds of functionally distinct, independently variable, causally
interacting substates characterise human minds? Are all cultures
the same? Are children and adults the same? Are there different
architectures for human-like minds? Where do other animals fit in.?
Session 4. Tea till exhaustion
Architectural requirements for affective states: desires,
emotions, pains, pleasures. Can machines have "raw feels"?
Can anything clear be extracted from the mess of current
thought about consciousness. Is there something to be explained?
What are the really hard unsolved philosophical problems
in AI and cognitive science. (The representation of arbitrary
perceivable shaps?)
Tutorial 6 - Knowledge Acquisition
----------
Dr Nigel Shadbolt
Univ of Nottingham
UK
We have no further details on Tutorial 6 - but one of us
(Caroline) attended the previous version and not only enjoyed
it but also thought it provided many useful knowledge
elicitation techniques. It is recommended for any one who needs
to acquire knowledge whether for an expert system task or some
other purpose.
We hope that this gives you a feel for the tutorials and that
you will enjoy one of them.
Caroline Knight & Steve Todd
AISB'91 Tutorial organisers
tutorials@hplb.hpl.hp.com
tutorials%hplb.uucp@ukc.ac.uk
tutorials@hplb.lb.hp.co.uk
All booking is being handled at leeds, contact:
aisb91@ai.leeds.ac.uk