Begin forwarded message:
From: Tom Lenaerts <[EMAIL PROTECTED]>
Date: Mon 13 Mar 2006 10:30:09 GMT+01:00
To: Carlos Gershenson <[EMAIL PROTECTED]>
Subject: can you also spread this around to interested people?
Advances in Bio-inspired Search and Optimization.
March 22, 2006.
Room D2.01 (Alouis Gerlo)
Vrije Universiteit Brussel
We would like to invite students, phd students, postdoctoral
researchers and professors with in an interest in bio-inspired
algorithms, combinatorial optimisation problems and biological
metaphors to participate in this seminar.
If you want to attend these presentations, send an email to Tom
Lenaerts so we can take into account the number of attendants. For
further information also send an email. These seminars are hosted
in co-organisation with the research lab CoMo (Vrije Universiteit
Brussel) and the FWO research group on Machine Learning for
Datamining and their Applications
10:30 Dr. Jano van Hemert
Dr. Van Hemert is application consultant at the National e-Science
institute in Edinburgh where he is involved the European project
GeneExpress.His research focus is on combinatorial problems and the
application of evolutionary techniques to solve them.
Not solving, but evolving hard combinatorial problems
Abstract:
We show how evolutionary computation can be used to acquire
difficult to solve combinatorial problem instances. The technique
is demonstrated on a number of well known domains of combinatorial
optimization, including binary constraint satisfaction and the
traveling salesman problem. Problem instances acquired through this
technique are more difficult than ones found in popular benchmarks.
We analyze these evolved instances with the aim to explain their
difficulty in terms of structural properties, thereby exposing the
weaknesses of corresponding algorithms.
11:20 Dr. Thomas Stuetzle
Dr. Stuetzle is chercheur qualife with the FNRS and is working at
IRIDIA, a well-known interdisciplinary research group at the
Universite Libre de Bruxelles. His research focus is on Stochastic
Local Search and optimization. He recently published two books on
Ant colony optimization and the principles of Stochastic Local Search.
Stochastic local search algorithms: Techniques and recent trends
Abstract:
Stochastic local search (SLS) algorithms like iterative
improvement, simulated annealing, iterated local search, ant colony
optimization or evolutionary algorithms are amongst the most
prominent and successful techniques for solving computationally
hard problems in many areas of Computer Science and Operations
Research like constraint optimization. In this talk we give a short
overview of the field of SLS algorithms with a special emphasis on
the main techniques available in SLS. The talk will then focus on
three example techniques that are important ingredients in a step
towards are more systematic development of SLS algorithms. These
three techniques are (i) iterated local search as a simple and
powerful, general-purpose SLS method, (ii) the usage of run-
timedistributions as a core tool for the experimental analysis of
SLS algorithms, and (iii) the usage of racing algorithms, in
particular F-races, for the automatisation of the design and tuning
of SLS algorithms. We end the talk by identifying the contributions
these techniques can give towards an engineering-based approach for
SLS algorithms.
12:10 Dr. Richard Watson
Dr. Watson is senior lecturer in the School of Electronics and
Computer Science at the University of Southampton. His recent
research focuses on exchange between evolutionary computation and
evolutionary biology. He recently published a book with MIT press
titled Compositional evolution: the impact of sex, symbiosis and
modularity on the gradualist framework of evolution.
Compositional Evolution: The impact of Sex, Symbiosis and
Modularity on The Gradualist Framework of Evolution
Abstract:
Conventionally, evolution by natural selection is almost
inseparable from the notion of accumulating successive slight
variations. Notions of gradualism are deeply ingrained in
evolutionary thought and strongly influence our ideas about what
kinds of systems are evolvable and which are not. The very idea of
large adaptive genetic changes is considered un-evolutionary.
Indeed, attacking the plausibility of a linear gradual succession
of proto-systems is a common form of criticism for evolutionary
theory from Paley (1802) to Behe (1996), and asserting the
plausibility of a gradual succession of proto-systems is a common
form of defense (Dawkins 1996). But is gradualism the only valid
framework for evolution by natural selection? Darwin's masterful
contribution was to provide an algorithmic model of how adaptation
may take place in biological systems. However, the hill-climbing
algorithm that he described - the sequential linear accumulation of
successive slight modifications - is only one algorithmic
possibility. In this work I show that compositional evolution,
utilizing mechanisms such as sexual recombination and symbiotic
encapsulation, do not fit within the hill-climbing framework and
provide a fundamental distinct form of adaptation process allied to
divide-and-conquer problem decomposition. Since compositional
evolution does not operate within the hill-climbing framework,
gradualist notions of evolutionary difficulty do not apply.
Accordingly, I am able to show that certain kinds of complex
systems are in principle easily evolvable under compositional
evolution despite the absence of a gradual succession of proto-
systems.
----------------------------------------------------------------------
---------------------------
Tom Lenaerts (tlenaert at ulb.ac.be)
+Postdoc Researcher @ SWITCH, Flemish Interuniversity Institute for
Biotechnology
+Guest Professor @ Department of Computer Science-Vrije Uiversiteit
Brussel-Belgium
http://switch.vub.ac.be/~tlenaert/
Carlos Gershenson...
Centrum Leo Apostel, Vrije Universiteit Brussel
Krijgskundestraat 33. B-1160 Brussels, Belgium
http://homepages.vub.ac.be/~cgershen/
“To know your limits you need to go beyond them”
