Analysis and Design of Representations and Operators (ADoRo'2003)

Franz Rothlauf and Dirk Thierens

Duration: Half Day

Successful and efficient use of evolutionary algorithms (EAs) depends on the choice of the genotype representation and the genetic operators.
These choices cannot been made independently of each other. The question whether a certain representation leads to better performing EAs than an alternative representation, can only be answered when the operators applied are taken into consideration. The reverse is also true: deciding between alternative operators is only meaningful for a given representation.
The application of different types of evolutionary search operators like, for example, mutation or crossover, should result in offspring that inherit some useful information from their parents. If a specific operator-represenation combination does not guarantee this, EAs turn into random search and guided search becomes impossible.
Despite the importance of choosing proper representation-operator combinations on the performance of EAs, little general applicable theory and knowledge is available to help understanding and guiding the construction of successful and efficient EAs. The purpose of the workshop is to provide a platform where these issues can be discussed. Relevant topics are (but not limited to):
- theoretical and empirical properties of representations and/or operators
- predictive performance measures for evaluating representation and operator choices
- redundant versus non-redundant genotype coding
- high-locality versus low-locality representations
- search space bias of representations and operators
- promising directions of future research

Dr. Franz Rothlauf
Lehrstuhl für ABWL und Wirtschaftsinformatik
Universität Mannheim
Schloss, Zimmer S134
D-68131 Mannheim
tel.: ++49 621/181-1689
fax: ++49 21/181-1471
e-Mail: rothlauf@uni-mannheim.de

Dr. Dirk Thierens
Institute of Information & Computing Sciences,
Utrecht University, Padualaan 14, 3584 CH Utrecht
The Netherlands
phone: +31-30-2534031
fax: +31-30-2513791
dirk.thierens@cs.uu.nl
http://www.cs.uu.nl/~dirk

Further details

Application of Hybrid Evolutionary Algorithms to NP-complete Problems

Francisco Baptista Pereira, Ernesto Costa, Günther Raidl

Duration: Half Day

This workshop will focus on the application of hybrid Evolutionary Computation (EC) techniques to NP-complete problems. The NP-complete decision class includes decision problems for which answers can be verified for correctness by an algorithm whose run time is polynomial in the size of the input. Moreover, if we obtain a polynomial algorithm that solves any NP complete problem, then it can be used to solve all other NP-problems quickly (i.e., in polynomial time).
There are many examples of problems belonging to this class, such as the boolean satisfiability (SAT), clique, decision trees, graph partitioning or Hamiltonian circuits. Moreover, optimisation problems belonging to the NP-hard complexity class, such as the travelling salesperson or bin packing can easily be restated in terms of a decision version. As an example, the optimisation question "What is the shortest tour?" which is NP-hard, can be restated as the NP-complete decision problem "Is there a tour length less than K?".
In the past few years there has been several attempts to apply evolutionary algorithms to NP-complete problems. Some good results have been achieved, showing that, at least in particular situations, it is possible to apply EC techniques to problems belonging to this class. Additionally, a number of difficulties have been identified, making the application of such algorithms to NP-complete problems a great challenge. Examples of difficulties referred by different researchers are the choice of a suitable representation, the design of efficient operators or the selection of a method to deal with constraints (which might be directly related to the choice of an appropriate fitness function).
With the purpose of improving the efficiency of search, several researchers developed hybrid architectures where evolutionary techniques are combined with some classical methods usually employed in combinatorial optimisation.
The reason for this merging is clear: EC algorithms are stochastic optimisation methods and, when searching for solutions in difficult problems (like those belonging to the NP-complete class), they might benefit from hybridising with classical techniques.
Adopting several current approaches as a starting point, this workshop aims to promote a widespread discussion about this topic and, most important, to analyse if it is possible to develop new hybrid architectures that perform better than today's methods. The workshop considers hybridisation in a general sense. This way, the combination between EC algorithms and classical methods includes the following possibilities (although not< limited to them):
- Hybridisation with exact techniques: Some classical exact methods, such
as branch and bound, dynamic programming or linear programming have been
applied to several NP-complete problems. The difficulty with these techniques is that most of them suffer from the scalability problem (in large instances they usually have severe difficulties). Nevertheless, EC algorithms might benefit from hybridising with this class of methods.
- Hybridisation with approximation algorithms: Approximation algorithms are methods that provide a guarantee on the quality of the solutions obtained.
There are many such methods, which were designed to specific NP-complete problems. Just like in the previous point, it is possible that merging an EC approach with an approximation algorithm helps to improve search performance.
- Using EC algorithms inside exact techniques: an evolutionary algorithm can be used to obtain good starting points for exact methods (for example, it can determine good bounds for a branch and bound strategy). This assistance might speed up the method and/or allow it to address larger instances of complex problems.
- One crucial question that arises from this debate is how well an EC algorithm (either used alone or merged with other techniques) might work on NP-complete problems with different characteristics. We consider that important knowledge could arise if researchers share the experience gained when trying to develop efficient algorithms to address such difficult problems. A few general themes that we consider are worth discussing are:
- What are the strengths (and weaknesses) of today's EC-based approaches?
How do they compare to other techniques that are also applied in such problems?
- Is there any particular kind of NP-complete problems for which EC
algorithms are particularly suited?

Francisco Baptista Pereira
Centro de Informática e Sistemas da Universidade de Coimbra,
Departamento de Engenharia Informática, Universidade de Coimbra Polo II,
3030 Coimbra, Portugal
Phone: +351 239790000, Fax: +351 239701266
Email: xico@dei.uc.pt

Ernesto Costa
Centro de Informática e Sistemas da Universidade de Coimbra,
Departamento de Engenharia Informática, Universidade de Coimbra Polo II,
3030 Coimbra, Portugal
Phone: +351 239790000, Fax: +351 239701266
Email: ernesto@dei.uc.pt

Günther Raidl
Institute of Computer Graphics and Algorithms, Vienna University of
Technology, Vienna, Austria.
Phone: + 43 (1)58801-18616, Fax: +43 (1)58801-18699
raidl@ads.tuwien.ac.at

Further details

Biological Applications for Genetic and Evolutionary Computation (BioGEC'2003)

Wolfgang Banzhaf and James Foster

Duration: Half Day

The field of Genetic and Evolutionary Computation has greatly benefited by borrowing ideas from Biology. Recently, it has become clear that GEC can help solve biological problems, and thereby to "repay the debt".
It is also becoming apparent that the computer itself can be used as a model organism with which to study evolutionary processes in nature.
This workshop is intended to raise the attention of GEC practitioners to interesting and challenging biological questions. We hope to bring together biologists and computer scientists for an exchange of ideas on problems and computational problem solving techniques in Biology.

Relevant topics include (but are not limited to):
- Data mining biological data repositories
- Sequence alignment
- Phylogenetic reconstruction
- Gene expression and regulation, alternate splicing
- Functional diversification through gene duplication and exon shuffling
- Structure Prediction for biological molecules (structural genomics and proteomics)
- Network reconstruction for developmental, expression, metabolism, catalysis, etc.
- Dynamical system approaches to biological systems
- Simulation of cells, viruses, organisms, and ecologies

Wolfgang Banzhaf
Department of Computer Science
Informatik 11
University of Dortmund
Joseph-von-Fraunhofer-Str. 20
44227 Dortmund,
GERMANY
email: banzhaf@cs.uni-dortmund.de
tel: +49-(0)231-9700-953
fax: +49-(0)231-9700-959
www: http://ls11-www.cs.uni-dortmund.de/people/banzhaf

James A. Foster
Department of Computer Science
University of Idaho
Moscow, ID 83844-1010
USA
email: foster@cs.uidaho.edu
tel: 208.885.7062
fax: 208 885-9052 (fax)
www: http://www.cs.uidaho.edu/~foster

Further details

Hardware Evolutionary Algorithms and Evolvable
Hardware (HEAEH 2003)

John C. Gallagher

Duration: Half Day

The emerging field of Evolvable Hardware (EH) has received a great deal of attention in recent years. Many EH researchers share the intention of building EH machines that combine an evolutionary algorithm (EA) engine with reconfigurable hardware into a single physical device that can evolve
on line to adapt to damage or unforeseen mission changes. This workshop will focus on "Hardware EAs" (HEAs) specifically designed for direct hardware implementation. We will discuss a taxonomy of HEA< and a number of HEA metrics, including speed, size, and search efficacy.
We will also discuss specific field programmable gate array (FPGA) and custom very large scale integration (VLSI) implementations of a number of specific algorithms.

Dr. John C. Gallagher
Wright State University
Dayton, OH 45435-0001
(937) 775-3929 [voice]
(937) 775-5133 [fax]
jgallagh@cs.wright.edu
http://gozer.cs.wright.edu/

Further details

Grammatical Evolution Workshop (GEWS'2003)

Michael O'Neill and Conor Ryan

Duration: Half Day

Following on from the success of the first Grammatical Evolution Workshop (GEWS 2002) held at GECCO 2002, the 2nd Grammatical Evolution Workshop will be held at GECCO 2003.

Grammatical Evolution (GE) is an automatic programming system that can evolve programs in an arbitrary language from a binary string. GE adopts a genotype-phenotype mapping process taking as input a grammar that describes the syntax of the evolved program. In addition to the grammar, the search algorithm (the standard has been a variable-length genetic algorithm) is also a 'plug-in' component of the system.
The workshop will address all aspects of GE including foundations, extensions, analysis and applications.

Michael O'Neill
Dept. of Computer Science & Information Systems
University of Limerick
Ireland
Email: michael.oneill@ul.ie
Tel: +353-61-213542
Fax: +353-61-202734

Conor Ryan
Dept. of Computer Science & Information Systems
University of Limerick
Ireland
Email: conor.ryan@ul.ie
Tel: +353-61-202755
Fax: +353-61-202734

[Further details]

Interactive Evolutionary Search and Exploration Systems

Ian Parmee

Duration: Half Day

In terms of design and decision-making, there is a role for evolutionary computation for optimal information gathering. A major advantage of population-based search techniques relates to their capability as powerful search and exploration algorithms that can provide diverse, interesting and potentially competitive solutions. Such solutions can provide information to the user which supports a better understanding of the problem domain and helps to define best directions for future investigation. This capability is extremely important when operating within ill-defined and uncertain decision-making environments where initial fitness functions are largely conceptual and the primary task is to improve definition and increase confidence. Information gained from initial search utilising conceptual models supports their development by the user in an iterative, interactive EC environment.
Although the development of such systems is ambitious, the requirement for such design and decision-making support is universal. It is difficult to think of any technology other than EC that can provide the level of underlying search and exploration required across ill-defined, uncertain problem spaces
Professor Ian Parmee,
Faculty of Computing, Engineering and Mathematical Science,
University of the West of England,
Coldharbour Lane,
Bristol, BS16 1QY,
UK.
Email: Ian.Parmee@uwe.ac.uk
Fax: ++44 117 3443155

Further details

International Workshop on Learning Classifier Systems

Wolfgang Stolzmann, Pier-Luca Lanzi, Stewart Wilson

Duration: Full Day

Submissions are invited that discuss recent developments in learning classifier systems research. All submitted papers will be peer reviewed by at least two members of the program committee. All accepted papers must be presented at the workshop by at least one of the authors and will be published in post-workshop proceedings. Papers should not be longer than twenty pages (Springer LNCS/LNAI style), including title page, figures, and bibliography.
For submission of longer papers please contact the members of the organizing committee.

Wolfgang Stolzmann
DaimlerChrysler AG
Research & Technology
Cognition and Robotics
Alt-Moabit 96A
D-10559 Berlin, Germany
Email: Wolfgang.Stolzmann@web.de


Pier-Luca Lanzi
Dipartimento di Elettronica e Informazione
Politecnico di Milano
Piazza Leonardo da Vinci, 32
I-20133 Milano, Italy
Tel: +39-02-2399-3472
Fax +39-02-2399-3411
Email: pierluca.lanzi@polimi.it

Stewart Wilson,
Prediction Dynamics,
Concord,
MA 01742,
USA
Email: wilson@prediction-dynamics.com

Further details

Learning, Adaptation, and Approximation in Evolutionary Computation

Sibylle Mueller, Petros Koumoutsakos, Marc Schoenauer
Yaochu Jin, Sushil Louis, and Khaled Rasheed

Duration: Full Day

A merger of the workshop on Learning and Adaptation in EC and the workshop
on Learning and Approximation in EC

In this workshop, we will bring together researchers from fields in machine learning and evolutionary optimization to discuss how the combination of learning, adaptation, and approximation can improve the efficiency of evolutionary algorithms.
Examples include:

- off-line and on-line learning for approximate model construction,
- off-line and on-line learning for performance improvement,
- step size adaptation techniques for evolution strategies,
- individual learning that guides evolution (Baldwin effect),
- self-organization and dimensionality reduction for evolving populations,
- domain knowledge extraction and reuse,
- evolution control and model management in evolutionary computation,
- multi-level evolutionary optimization,
- learning in multi-objective evolutionary optimization,
- fitness estimation in noisy environment,
- comparison of different modeling methods, such as neural networks, response surface, Gaussian processes, least squares methods, and probabilistic models for evolutionary computation,
- comparison of different sampling techniques for on-line and off-line learning.

Dr. Sibylle Mueller
Institute of Computational Science
Swiss Federal Institute of Technology (ETH) Zuerich
ETH Zentrum, Hirschengraben 84, HRS H3
CH-8092 Zuerich, Switzerland
Tel : +41-1-632 6827
Fax: +41-1-632 1703
Email: muellers@inf.ethz.ch

Prof. Petros Koumoutsakos
Institute of Computational Science
Swiss Federal Institute of Technology (ETH) Zuerich
ETH Zentrum, Hirschengraben 84, HRS H3
CH-8092 Zuerich, Switzerland
Tel : +41-1-632 5258
Fax: +41-1-632 1703
Email: petros@inf.ethz.ch

Prof. Marc Schoenauer
Projet FRACTALES - I.N.R.I.A. Rocquencourt
B.P. 105 - 78153 LE CHESNAY Cedex - France
Tel : +33-139 63 50 87
Fax : +33-139 63 59 95
Email: Marc.Schoenauer@inria.fr

Dr. Yaochu Jin
Future Technology Research
Honda R&D Europe
Carl-Legien-Str. 30
63073 Offenbach/Main, Germany
Tel: +49-69-89011735
Fax: +49-69-89011749
Email: yaochu.jin@de.hrdeu.com

Prof. Sushil Louis
Department of Computer Science/171
University of Nevada, Reno, NV 89557-0148
Tel: +1-775-784 1877
Fax: +1-775-784 4315
http://www.cs.unr.edu/~sushil
Email: sushil@cs.unr.edu

Prof. Khaled Rasheed
Computer Science Department
The University of Georgia
Athens, GA 30602-7404
Tel: +1-706-542 3444
Fax: +1-706-542 2966
http://www.cs.uga.edu/~khaled
Email: khaled@cs.uga.edu

Further details

Workshop on Memetic Algorithms 2003 (WOMA-IV)

Peter Merz, William E. Hart, Natalio Krasnogor, Jim E. Smith

Duration: Full Day

The next international Workshop on Memetic Algorithms (WOMA-IV), will be the fourth in a series of workshops dedicated exclusively to Memetic Algorithms. The WOMA series is a forum where the international community of researchers, practitioners and vendors, that work on aspects related to memetic algorithms, can engage in fruitful discussions, learning and contribute to the advancement of our field.
Memetic algorithms (MAs) are evolutionary algorithms (EAs) that apply a separate local search process to refine individuals (e.g. improve their fitness by hill-climbing). These methods are inspired by models of adaptation in natural systems that combine evolutionary adaptation of populations of individuals with individual learning within a lifetime. Additionally, MAs are inspired by Richard Dawkin's concept of a meme, which represents a unit of cultural evolution that can exhibit local refinement. Thus a memetic model of adaptation exhibits the plasticity of individuals that a strictly genetic model fails to capture. Under different contexts and situations, MAs are also known as hybrid EAs, genetic local searchers, Baldwinian EAs, Lamarkian
EAs, etc.
From an optimization point of view, MAs are hybrid EAs that combine global and local search by using an EA to perform exploration while the local search method performs exploitation. Combining global and local search is a strategy used by many successful optimization approaches, and MAs have in fact been recognized as a powerful algorithmic paradigm for evolutionary computing. In particular, the relative advantage of MAs over EAs is quite consistent on complex search spaces.
It is the goal of this new edition of the workshop to push forward our understanding of both the theory and the deployment of MA. The themes of the workshop include (but are not limited to):
- Memetic algorithms applications:
scheduling, transport, logistic, network optimization, process optimization space craft trajectory optimization, bioinformatics, planning, timetabling, evolvable hardware and hardware design, robotics, telecommunications, mechanical and structural engineering
- Memetic algorithms theory
- Theory of MAs and memetics
- Software engineering issues
- Kolmogorov, Computational and PLS Complexity issues
- Convergence of MAs
- Competent MAs
- Distributed/Parallel MAs
- Theoretical/Experimental comparisons/integration with other soft
techniques, e.g., exact methods, expert systems, simulated annealing, knowledge based systems, heuristic search, tabu search, ant colony optimization, genetic programming, etc.
- MAs for Multiobjective optimization, discrete and continuous optimization
- MAs for mixed domains
- MAs for optimisation of non-stationary problems
- Frameworks for describing and classifying MAs
- Practical guidelines to combine local search and EAs
- Scalability of MAs
- New MA architectures
- MA performance predictions
- Landscape analysis

Peter Merz
Department of Computer Science (WSI)
University of Tübingen
Sand 1,
D-72076 Tübingen, Germany.
Tel: (+49) 7071 / 29 77175
Fax: (+49) 7071 / 29 5091
Email: pmerz@informatik.uni-tuebingen.de


William E. Hart
Optimization/Uncertainty Estimation Dept (9211), MS 1110
P.O. Box 5800, Sandia National Labs
Albuquerque, NM 87185-1110
Tel: (505) 844-2217
Fax: (505) 845-7442
Email: wehart@cs.sandia.gov

Natalio Krasnogor
Automated Scheduling, Optimisation and Planning Research Group
School of Computer Science and Information Technology
University of Nottingham
University Park, Nottingham NG7 2RD
United Kingdom
Tel: (44) 115 9513477
Email: Natalio.Krasnogor@nottingham.ac.uk

Jim E. Smith
Intelligent Computer Systems Centre
Faculty of Computer Studies and Mathematics
University of the West of England
Coldarbour Lane,
Bristol, BS16 1QY
United Kingdom.
Tel: +44 (0) 117 3443161
Fax: +44 (0) 117 9750416
Email:James.Smith@uwe.ac.uk

Further details