| |
More invited speakers will appear here soon
Constraint programming is a simple paradigm. One models a problem by
stating constraints on acceptable solutions and then uses a general
purpose search algorithm to find a solution which satisfies the
constraints. The approach is powerful and has led to state-of-the-art
solutions to many important problems, including instruction scheduling
in compilers and model checking in formal verification. In this talk,
I will review the fundamental techniques of constraint programming and
then discuss a long-standing research program that I and others have
pursued on determining which backtracking search algorithms are the best.
I received my PhD in 1990 from the University of Waterloo and at that
time joined the faculty at the University of Alberta. Ten years later
I returned to Waterloo. I have co-authored four conference papers which
have won awards, and have served on or am serving on the editorial boards
of three journals. My research interests are in constraint programming
and its application to scheduling, sequencing, and planning.
The computer age's centralized mindset has successfully produced machines that have changed our lives. But today's computer isn't the only possible tool for computing. One alternative is swarm intelligence. Forget centralization and control. Forget programming. Forget the concept of a big, omniscient computer. Think of a hive, or an anthill. Social insect colonies aren't centrally controlled; they're composed of thousands or even millions of insects with limited cognitive repertoires. Individually, one insect can't do much, but collectively, social insects can achieve great things build a nest, forage for food, take care of the brood, allocate labor, and so on. The collective intelligence of social insects, swarm intelligence, offers a powerful new model for computing, based on local reinforcement and global dissipation. Researchers have successfully applied the swarm approach to a number of real optimization and control problems, from factory scheduling to telecommunications-network routing. The ant approach is always the one with the most flexibility in response to changing conditions. Beyond optimization and control, swarm intelligence provides an entirely new way of thinking about and harnessing distributed computing.
Eric Bonabeau is the founder and chief scientist of Icosystem Corporation, a Boston-based idea incubator that invents novel business models using a combination of simulation and evolutionary computation. Prior to his current position, Dr. Bonabeau was the CEO of Eurobios, a consultancy applying the science of complex adaptive systems to business issues. He has been a research director for CNET France Telecom, an R&D engineer at Cadence Design Systems (in Lowell, MA, USA), and the Interval Research Fellow at the Santa Fe Institute. He is the author of more than one hundred science articles and three books (Intelligence Collective, HermC's, 1994; Swarm Intelligence, Oxford University Press, 1999; and Self-Organization in Biological Systems, Princeton University Press, 2001). Dr. Bonabeau is co-editor-in-chief of Advances in Complex Systems and a member of the editorial and scientific committees of more than twenty-five international journals and conferences.
The explosive growth of genomic data quickly makes many
classical bioinformatics software obsolete and presents
new and challenging foundamental and practical questions.
Simple heuristics that worked beautifully when
data size was small are no longer sufficient.
For example, it takes years for Blast to compare
the human genome and the mouse genome (on a modern computer).
New ideas, sophisticated algorithm design,
and foundamental research in computer science
and AI are urgently needed in the new generation
scalable and high quality bioinformatics software.
This will be demonstrated by several problems, and some
solutions, including homology search, genome comparison, multiple
alignment, protein structure prediction, mass spectrometry de novo
sequencing, and phylogeny construction.
Ming Li is a professor of Computer Science at the University of
California Santa Barbara. He is a recipient of Canada's
E.W.R. Steacie Followship Award in 1996, and the 2001 Killam Fellowship.
He is a coauthor of the book "An Introduction to Kolmogorov Complexity and
Its Applications" (Springer-Verlag, 1993, 2nd Edition, 1997).
He currently serves on the editorial boards of Journal of Computer and
System Sciences, Information and Computation, Journal of Combinatorial
Optimization, and International Journal of Foundation of Computer
Science. He is a co-managing editor of the new Journal of Bioinformatics
and Computatational Biology. His main research area is Bioinformatics.
The main task of a soccer player is to score goals. However, there are moments in life when questions like the following become relevant: Is the ball I am seeing a hallucination or is it real? Should I revise my beliefs about where the ball is? And if so, what is the next action I should execute? Would this action be to the benefit of my team? In the talk I will address these questions and show how one can create a successful robotic soccer team by giving the right answers.
Bernhard Nebel received his Ph.D. (Dr. rer. nat.) from the University
of Saarland in 1989. From 1993 to 1996 he held an Associate Professor
position (C3) at the University of Ulm. Since 1996 he is Professor at
Albert-Ludwigs-Universität Freiburg. Among other professional
services, he served as the Program Co-chair for the 3rd International
Conference on Principles of Knowledge Representation and Reasoning
(KR'92) and as the Program Chair for the 17th International Joint
Conference on Artificial Intelligence (IJCAI'01). In addition, he is a
member of the editorial board of AI Communication and he is the
Research Note editor of Artificial Intelligence. His main research
interests are knowledge representation and reasoning, planning, and
robotics with an emphasis on robotic soccer.
Imitation and social learning provide mechanisms by which social animals,
robots, or software agents are able to acquire competencies by observing
the behaviour of others. Organisms with different types of embodiment,
different histories, different skills and experience might learn from one
another while sharing a context, without having to solve every problem
themselves for the first time, but may harness the experience of others.
Such transmission of behaviours and skills contribute the bases for teaching
and the evolution of cultures in humans and other social animals. Without
having to be explicitly programmed and without having to undertake individual
learning, an organism may acquire useful new skills. We are developing
general methods to do this for constructed agents, such as robots and
software agents.
These aspects of intelligence are distinct from those that focus on
individualized learning commonly studied in artificial intelligence, and
also distinct from non-individualized swarm intelligence in which individuals
are interchangeable and do not learn. The possibility of social learning
raises deep questions of perception and action, development and
communication in a shared context, and requires the articulation of important
issues such as the correspondence problem (attempting to construct a
correspondence between differently embodied organisms); what, who, when,
how and why to imitate, and how to judge success (metrics); and
distinctions such as learning to imitate vs. learning by imitation.
We will survey the background, problems and issues of social learning,
including foundational work, as well as some implementations with robotic
and artificial social worlds in which social learning takes place by
K. Dautenhahn, A. Billard, A. Alissandrakis, C. Nehaniv, Y. Demiris, and
others.
Chrystopher L. Nehaniv received his PhD from the University of California,
Berkeley, in 1992, and subsequently held research and academic positions
at Berkeley, Japan, Hungary, and the U.K. He is currently Professor of
Mathematical and Evolutionary Computer Sciences with the Adaptive Systems
Research Group in the Faculty of Engineering and Information Sciences at
the University of Hertfordshire, Hatfield, England. He is author of numerous
scientific publications in areas ranging from theoretical computer science
to biology, has edited numerous journal issues and books including
Mathematical and Computational Biology (AMS, 1999), Computation for
Metaphors, Analogy, and Agents (Springer LNAI, 1999), Imitation in Animals
and Artifacts (MIT Press, 2002; with K. Dautenhahn), and others.
Prof. Nehaniv is Director of U.K. Engineering and Physical Sciences Research
Council Network on Evolvability in Biological and Software Systems, and is
Associate Editor of the journal BioSystems.
Many research projects in AI use functional and logic programming languages.
While the most widely used such languages (Lisp and Prolog) are fairly old,
functional and logic languages that are purely declarative in style are a
relatively recent development. This talk will explain how the Mercury logic
programming language supports purely declarative programming, what the main
advantages of the purely declarative programming style are, and how its
drawbacks can be mitigated.
Zoltan Somogyi received his PhD in 1989 from the University of Melbourne for
research on parallel logic programming. After a detour working on
deductive databases and information retrieval systems, he returned to
his roots, founding the Mercury project in 1993. This project has since
grown to become one of the biggest in the department, attracting substantial
funding (from Microsoft as well as the ARC) and several industrial users.
The aim of the project was to build a new logic programming language,
Mercury, with two goals in mind:
- making the language purely declarative, to make all the nice theorems
proven by logic programming researchers over the years apply to the language;
- making the language suitable for programming in the large.
The project has achieved these objectives. It has also demonstrated that
a purely declarative logic programming language can be very practical,
as well as very fast; Mercury is the fastest logic programming language by far.
Its implementation reflects Zoltan's research interests and his recent work:
it comes with a wide array of program analyses and optimizations, as well as
a range of programming tools, including a very detailed program profiler
and an industrial-strength declarative debugger.
|
|
