The annual EuroGP series are the premier conferences in Europe devoted entirely to genetic programming. EuroGP conferences are always enjoyable and offer good opportunities for informal contact with fellow researchers in a friendly and relaxed setting, and this year will be held at the University of Coimbra, Portugal, in conjunction with EvoCOP2004 and EvoWorkshops2004. The EuroGP2004 conference is a mixture of oral presentations and poster sessions and ALL accepted papers appear as full papers in the proceedings published by Springer as part of EvoWorkshops2004 in the Lecture Notes in Computer Science series.

LNCS 3003, the proceedings for EuroGP2004, is now available online

EUROGP PROGRAMME DETAILS

CONFERENCE PAPERS:

Oral Presentations

Evaluation of chess position by modular neural network generated by genetic algorithm
Mathieu Autones, Aryel Beck, Phillippe Camacho, Nicolas Lassabe, Herve Luga, Franccois Scharffe

Abstract:
In this article we present our chess engine Tempo. One of the major difficulties for this type of program lies in the function for evaluating game positions. This function is composed of a large number of parameters which have to be determined and then adjusted. We propose an alternative which consists in replacing this function by an artificial neuron network (ANN). Without topological knowledge of this complex network, we use the evolutionist methods for its inception, thus enabling us to obtain, among other things, a modular network. Finally, we present our results: (i) reproduction of the XOR function which validates the method used and (ii) generation of an evaluation function

Coevolution of Algorithms and Deterministic Solutions of Equations in Free Groups
Richard F. Booth, Alexandre V. Borovik

Abstract:
We discuss the use of evolutionary algorithms for solving problems in combinatorial group theory, using a class of equations in free groups as a test bench. We find that, in this context, there seems to be a correlation between successful evolutionary algorithms and the existence of good deterministic algorithms. We also trace the convergence of co-evolution of the population of fitness functions to a deterministic solution.

Designing Optimal Combinational Digital Circuits Using a Multiple Logic Unit Processor
Sin Man Cheang, Kin Hong Lee, Kwong Sak Leung

Abstract:
Genetic Parallel Programming (GPP) is a novel Genetic Programming paradigm. The GPP Accelerating Phenomenon, i.e. parallel programs are easier to be evolved than sequential programs, opens up a new approach to evolve solution programs in parallel forms. Based on the GPP paradigm, we developed a combinational digital circuit learning system, the GPP+MLP system. An optimal Multiple Logic Unit Processor (MLP) is designed to evaluate genetic parallel programs. To show the effectiveness of the proposed GPP+MLP system, four multi-output Binary arithmetic circuits are used. Experimental results show that both the gate counts and the propagation gate delays of the evolved circuits are less than conventional designs. For example, in a 3-bit multiplier experiment, we obtained a combinational digital circuit with 26 two-input logic gates in 6 gate levels. It utilizes 4 gates less than a conventional design.

A Data Structure for Improved GP Analysis via Efficient Computation and Visualisation of Population Measures
Aniko Ekart, Steven Gustafson

Abstract:
Population measures for genetic programs are defined and analysed in an attempt to better understand the behaviour of genetic programming. Some measures are simple, but do not provide sufficient insight. The more meaningful ones are complex and take extra computation time. Here we present a unified view on the computation of population measures through an information hyper-tree (iTree). The iTree allows for a unified and efficient calculation of population measures via a basic tree traversal.

Boosting technique for Combining Cellular GP Classifiers
Gianluigi Folino, Clara Pizzuti, Giandomenico Spezzano

Abstract:
An extension of Cellular Genetic Programming for data classification with the boosting technique is presented and a comparison with the bagging-like majority voting approach is performed. The method is able to deal with large data sets that do not fit in main memory since each classifier is trained on a subset of the overall training data. Experiments showed that, by using a sample of reasonable size, the extension with these voting algorithms enhances classification accuracy at a much lower computational cost.

Co-evolving Faults to Improve the Fault Tolerance of Sorting Networks
Michael L. Harrison, James A. Foster

Abstract:
Co-evolving Faults to Improve the Fault Tolerance of Sorting Networks Fault tolerance is an important objective for circuit design, so it is natural to apply genetic programming techniques that are already being used for circuit design to enhance fault tolerance. We present preliminary evidence that co-evolving faults with circuits enhances the masking of faults in evolved circuits. Our test systems are sorting networks, since these are simple enough to analyze. We show that the overall impact of faults in an evolved sorting network can be reduced proportionally to the strength of co-evolutionary pressure.

Toward an Alternative Comparison between Different Genetic Programming Systems
Xuan Nguyen, Bob McKay, Daryl Essam, Hussein Abbass

Abstract:
In this paper, we use multi-objective techniques to compare different genetic programming systems, permitting our comparison to concentrate on the effect of representation and separate out the effects of different search space sizes and search algorithms. Experimental results are given, comparing the performance and search behavior of Tree Adjoining Grammar Guided Genetic Programming (TAG3P) and Standard Genetic Programming (GP) on some standard problems.

Lymphoma Cancer Classification Using Genetic Programming with SNR Features
Jin-Hyuk Hong, Sung Bae Cho

Abstract:
Lymphoma cancer classification with DNA microarray data is one of important problems in bioinformatics. Many machine learning techniques have been applied to the problem and produced valuable results. However the medical field requires not only a high-accuracy classifier, but also the in-depth analysis and understanding of classification rules obtained. Since gene expression data have thousands of features, it is nearly impossible to represent and understand their complex relationships directly. In this paper, we adopt the SNR (Signal-to-Noise Ratio) feature selection to reduce the dimensionality of the data, and then use genetic programming to generate cancer classification rules with the features. In the experimental results on Lymphoma cancer dataset, the proposed method yielded 96.6% test accuracy in average, and an excellent arithmetic classification rule set that classifies all the samples correctly is discovered by the proposed method.

A Practical Approach to Evolving Concurrent Programs
David Jackson

Abstract:
Although much research has been devoted to devising genetic programming systems that are capable of running the evolutionary process in parallel, thereby improving execution speed, comparatively little effort has been expended on evolving programs which are themselves inherently concurrent. A suggested reason for this is that the vast number of parallel execution paths that are open to exploration during the fitness evaluation of population members renders evolutionary computation prohibitively expensive. We have therefore investigated the potential for minimising this expense by using a far more limited exploration of the execution state space to guide evolution. The approach, involving the definition of sets of schedulings to enable a variety of execution interleavings to be specified, has been applied to the classic ‘dining philosophers’ problem, and has been found to evolve solutions that are as good as those created by human programmers.

Evolutionary Induction of Grammar Systems for Multi-agent Cooperation
Clayton M. Johnson, James Farrell

Abstract:
We propose and describe a minimal cooperative problem that captures essential features of cooperative behavior and permits detailed study of the mechanisms involved. We characterize this problem as one of language generation by cooperating grammars, and present initial results for language induction by pairs of right-linear grammars using grammatically based genetic programming. Populations of cooperating grammar systems were found to induce grammars for regular languages more rapidly than non-cooperating controls. Cooperation also resulted in greater absolute accuracy in the steady state, even though the control performance exceeded that of prior results for the induction of regular languages by a genetic algorithm.

Genetic Programming Applied to Mixed Integer Programming
Konstantinos Kostikas, Charalambos Fragakis

Abstract:
We present the application of Genetic Programming (GP) in Branch and Bound (B&B) based Mixed Integer Linear Programming (MIP). The hybrid architecture introduced employs GP as a node selection expression generator: a GP run, embedded into the B&B process, exploits the characteristics of the particular MIP problem being solved, evolving a problem-specific node selection method. The evolved method replaces the default one for the rest of the B&B. The hybrid approach outperforms depth-first and breadth-first search, and compares well with the advanced Best Projection method.

Efficient Crossover in the GAuGE System
Miguel Nicolau, Conor Ryan

Abstract:
This paper presents a series of context-preserving crossover operators for the GAuGE system. These operators have been designed to respect the representation of genotype strings in GAuGE, thereby making sensible changes at the genotypic level. Results on a set of problems suggest that some of these operators can improve the maintenance and propagation of building blocks in GAuGE, as well as its scalability, and could be of use to other systems using structural evolving genomes.

Grammatical Evolution by Grammatical Evolution: The Evolution of Grammar and Genetic Code
Michael O'Neill, Conor Ryan

Abstract:
This study examines the possibility of evolving the grammar that grammatical evolution uses to specify the construction of a syntactically correct solution. As the grammar dictates the space of symbols that can be used in a solution, its evolution represents the evolution of the genetic code itself. Results provide evidence to show that the co-evolution of grammar and genetic code with a solution using grammatical evolution is a viable approach.

Constrained Molecular Dynamics as a Search and Optimization Tool
Riccardo Poli, Christopher R. Stephens

Abstract:
In this paper we consider a new class of search and optimization algorithms inspired by molecular dynamics simulations in physics.

On the performance of Genetic Operators the Random Key Representation
Eoin Ryan, Atif Azad, Conor Ryan

Abstract:
Many evolutionary systems have been developed that solve various specific scheduling problems. In this work, one such permutation based system, which uses a linear GP type Genotype to Phenotype Mapping (GPM), known as the Random Key Genetic Algorithm is investigated. The role standard mutation plays in this representation is analysed formally and is shown to be extremely disruptive. To ensure small fixed sized changes in the phenotype a swap mutation operator is suggested for this representation. An empirical investigation reveals that swap mutation outperforms the standard mutation to solve a hard deceptive problem even without the use of crossover. Swap mutation is also used in conjunction with different crossover operators and significant boost has been observed in the performance especially in the case of headless chicken crossover that produced surprising results.

Analysis of GP Improvement Techniques over the Real-World Inverse Problem of Ocean Color
Grégory Valigiani, Cyril Fonlupt, Pierre Collet

Abstract:
This paper is a follow-up of Maarten Keijzer's award-winning EUROGP'03 paper [Keij03], that suggests using Interval Arithmetic (IA) and Linear Scaling (LS) in Genetic Programming algorithms. The ideas exposed in this paper were so nice that it was decided to experiment with them on a real-world problem on which the LIL research team had some experience and results with: the Ocean Color Inverse Problem. After extensive testing of IA, LS as well as a progressive learning method using thresholds (T), results seem to show that functions evolved with GP algorithms that do not implement IA may output erroneous values outside the learning set, while LS and T methods produce solutions with a greater generalisation error. A simple and apparently harmless improvement over standard GP is also proposed, that consists in weighting operands of + and - operators.

Evolution and Acquisition of Modules in Cartesian Genetic Programming
James Alfred Walker, Julian Francis Miller

Abstract:
The paper presents for the first time automatic module acquisition and evolution within the graph based Cartesian Genetic Programming method. The method has been tested on a set of even parity problems and compared with Cartesian Genetic Programming without modules. Results are given that show that the new modular method evolves solutions up to 20 times quicker than the original non-modular method and that the speedup is more pronounced on larger problems. Analysis of some of the evolved modules shows that often they are lower order parity functions. Prospects for further improvement of the method are discussed.

How to Choose Appropriate Function Sets for GP
Wang Gang, Terence Soule

Abstract:
The choice of functions in a genetic program can have a significant effect on the GP's performance, but there have been no systematic studies of how to select functions to optimize performance. In this paper, we investigate how to choose appropriate function sets for general genetic programming problems. For each problem multiple functions sets are tested. The results show that functions can be classified into function groups of equivalent functions. The most appropriate function set for a problem is one that is optimally diverse; a set that includes one function from each function group.

Improving Grammer Based Evolution Algorithms via Attributed Derivation Trees
Szilvia Zvada, Robert Vanyi

Abstract:
Using Genetic Programming difficult optimization problems can be solved, even if the candidate solutions are complex objects. In such cases, it is a costly procedure to correct or replace the invalid individuals that may appear during the evolutionary process. Instead of such post-processing, context-free grammars can be used to describe the syntax of valid solutions, and the algorithm can be modified to work on derivation trees, such that it does not generate invalid individuals. Although tree operators have the advantage of good parameterizability, it is not trivial to construct them correctly and efficiently. In this paper an already existing method for derivation tree evolution and its extension towards attributed derivation trees are discussed. As the result of this extension the operators are not only faster but they are easy to parameterize, moreover the algorithm is better guided, thus it can converge faster.

Posters

Evolved Matrix Operations for Post-Processing Protein Secondary Structure Predictions
Varun Aggarwal, Robert MacCallum

Abstract:
Predicting the three-dimensional structure of proteins is a hard problem, so many have opted instead to predict the secondary structural state (usually helix, strand or coil) of each amino acid
residue. This should be an easier task, but it now seems that a ceiling of around 76% per-residue three-state accuracy has been reached. Further improvements will require the correct processing of so-called "long-range information". We present a novel application of genetic programming to evolve highlevel matrix operations to post-process secondary structure prediction probabilities produced by the popular, state-of-the-art neural networkbased PSIPRED by David Jones. We show that global and long-range information may be used to increase three-state accuracy by at least 0.26 percentage points - a small but statistically significant difference. This is on top of the 0.14 percentage point increase already made by PSIPRED's built-in filters.

Genetic Programming for Natural Language Parsing
Lourdes Araujo

Abstract:
The aim of this paper is to prove the effectiveness of the genetic programming approach in automatic parsing of sentences of real texts. Classical parsing methods are based on complete search techniques to find the different interpretations of a sentence. However, the size of the search space increases exponentially with the length of the sentence or text to be parsed and the size of the grammar, so that exhaustive search methods can fail to reach a solution in a reasonable time. This paper presents the implementation of a probabilistic bottom-up parser based on genetic programming which works with a population of partial parses, i.e. parses of sentence segments. The quality of the individuals is computed as a measure of its probability, which is obtained from the probability of the grammar rules and lexical tags involved in the parse. In the approach adopted herein, the size of the trees generated is limited by the length of the sentence. In this way, the size of the search space, determined by the size of the sentence to parse, the number of valid lexical tags for each words and specially by the size of the grammar, is also limited.

Comparing hybrid systems to design and optimize artificial neural networks
Pedro A. Castillo, Maribel G. Arenas, J.J. Merelo, Gustavo Romero, Fatima Rateb, Alberto Prieto

Abstract:
In this paper we conduct a comparative study between hybrid methods to optimize multilayer perceptrons: a model that optimizes the architecture and initial weights of multilayer perceptrons; a parallel approach to optimize the architecture and initial weights of multilayer perceptrons; a method that searches for the parameters of the training algorithm, and an approach for cooperative co-evolutionary optimization of multilayer perceptrons. Obtained results show that a co-evolutionary model obtains similar or better results than specialized approaches, needing much less training epochs and thus using much less simulation time.

An Evolutionary Algorithm for the Input-Output Block Assignment Problem
Kit Yan Chan, Terence C. Fogarty

Abstract:
In this paper, a procedure for system decompositon is developed for decentralized multivariable systems. Optimal input-output pairing techniques are used to rearrange a large multivariable\ system into a structure that is closer to the block-diagonal decentralized form. The problem is transformed into a block assignment problem. An evolutionary algorithm is developed to solve this hard IP problem. The result shows that the proposed algorithm is simple to implement and efficient to find the reasonable solution.

Genetic Programming for Subjective Fitness Function Identification
Dan Costelloe, Conor Ryan

Abstract:
This work addresses the common problem of modeling fitness functions for Interactive Evolutionary Systems. Such systems are necessarily slow because they need human interaction for the fundamental task of fitness allocation. The research presented here demonstrates that Genetic Programming can be used to learn subjective fitness functions from human subjects, using historical data from an Interactive Evolutionary system for producing pleasing drum patterns. The results indicate that GP is capable of performing symbolic regression even when the number of training cases is substantially less than the number of inputs.

Saving Effort in Parallel GP by means of Plagues
Francisco Fernández, Aida Martín

Abstract:
Recently, a new technique that allows Genetic Programming to save computing resources has been proposed. This technique was presented as a new operator acting on the population, and was called plague. By removing some individuals every generation, plague aims at compensating for the increase in size of individuals, thus saving computing time when looking for solutions. By means of some test problems, we show that the technique is also useful when employing a parallel version of GP, such as that based on the island model.

Sampling of Unique Structures and Behaviours in Genetic Programming
Steven Gustafson, Edmund K. Burke, Graham Kendall

Abstract:
This paper examines the sampling of unique structures and behaviours in genetic programming. A novel description of behaviour is used to better understand the solutions visited during genetic programming search. Results provide new insight about deception that can be used to improve the algorithm and demonstrate the capability of genetic programming to sample different large tree structures during the evolutionary process.

The Evolution of Concurrent Control Software Using Genetic Programming
John Hart, Martin Shepperd

Abstract:
Despite considerable progress in GP over the past 10 years, there are many outstanding challenges that need to be addressed before it will be widely deployed for developing useful software. In this paper we suggest a method for the automatic creation of concurrent control software using Linear Genetic Programming (LGP) and a `divide and conquer' approach. The method involves decomposing the whole problem into a multi-task solution with multiple inputs and multiple outputs -- similar to the process used to implement embedded control solutions. We describe the necessary architecture of typical embedded control systems and their relevance to this work, the software evolution scheme used and lastly demonstrate the technique for an embedded software problem, namely a washing machine controller.

Extending Grammatical Evolution to Evolve Digital Surfaces with Genr8
Martin Hemberg, Una-May O'Reilly

Abstract:
Genr8 is a surface design tool for architects. It uses a grammar-based generative growth model that produces surfaces with an organic quality. Grammatical Evolution is used to help the designer search the universe of possible surfaces. We describe how we have extended Grammatical Evolution, in a general manner, in order to handle the grammar used by Genr8.

Evolving Text Classifiers with Genetic Programming
Laurence Hirsch, Masoud Saeedi, Robin Hirsch

Abstract:
We describe a method for using Genetic Programming (GP) to evolve document classifiers. GP’s create regular expression type specifications consisting of particular sequences and patterns of N-Grams (character strings) and acquire fitness by producing expressions, which match documents in a particular category but do not match documents in any other category. Libraries of N-Gram patterns have been evolved against sets of pre-categorised training documents and are used to discriminate between new texts. We describe a basic set of functions and terminals and provide results from a categorisation task using the 20 Newsgroup data.

Automatic Synthesis of Instruction Decode Logic by Genetic Programming
David Jackson

Abstract:
On many modern computers, the processor control unit is microprogrammed rather than built directly in hardware. One of the tasks of the microcode is to decode machine-level instructions: for each such instruction, it must be ensured that control-flow is directed to the appropriate microprogram for emulating it. We have investigated the use of genetic programming for evolving this instruction decode logic. Success is highly dependent on the number of opcodes in the instruction set and their relationship to the conditional branch and shift instructions offered on the microarchitecture, but experimental results are promising.

Alternatives in Subtree Caching for Genetic Programming
Maarten Keijzer

Abstract:
This work examines a number of subtree caching mechanisms that are capable of adapting during the course of a run while maintaining a fixed size cache of already evaluated subtrees. A cache update and flush mechanism is introduced as well as the benefits of vectorized evaluation over the standard case-by-case evaluation method for interpreted genetic programming systems are discussed. The results show large benefits for the use of even very small subtree caches. One of the approaches studied here can be used as a simple add-on module to an existing genetic programming system, providing an opportunity to improve the runtime efficiency of such a system.

Structural Risk Minimization on Decision Trees Using An Evolutionary Multiobjective Optimization
DaeEun Kim

Abstract:
Inducing decision trees is a popular method in machine learning. The information gain computed for each attribute and its threshold helps finding a small number of rules for data classification. However, there has been little research on how many rules are appropriate for a given set of data. In this paper, an evolutionary multi-objective optimization approach with genetic programming will be applied to the data classification problem in order to find the
minimum error rate for each size of decision trees. Following structural risk minimization suggested by Vapnik, we can determine a desirable number of rules with the best generalization performance. A hierarchy of decision trees for classification performance can be provided and it is compared with C4.5 application.

Global Distributed Evolution of L-Systems Fractals
W. B. Langdon

Abstract:
Internet based parallel genetic programming (GP) creates fractal patterns like Koch's snow flake. Pfeiffer, http://www.cs.ucl.ac.uk/staff/W.Langdon/pfeiffer.html, by analogy with seed/embryo development, uses Lindenmayer grammars and LOGO style turtle graphics written in Javascript and Perl. 298 novel pictures were produced. Images are placed in animated snow globes (computerised snowstorms) by www web browsers anywhere on the planet. We discuss artificial life (Alife) evolving autonomous agents and virtual creatures in higher dimensions from a free format representation in the context of neutral networks, gene duplication and the evolution of higher order genetic operators.

Reusing Code in Genetic Programming
Edgar Galvan Lopez, Riccardo Poli, Carlos A. Coello Coello

Abstract:
In this paper we propose an approach to Genetic Programming based on reuse of code and we test our algorithm in the design of combinational logic circuits at the gate-level. The proposed algorithm is validated using examples taken from the evolvable hardware literature, and is compared against circuits produced by human designers, by Particle Swarm Optimization, by an n-cardinality GA and by Cartesian Genetic Programming.

Exploiting Reflection in Object Oriented Genetic Programming
Simon Lucas

Abstract:
Most programs currently written by humans are object-oriented ones. Two of the greatest benefits of object oriented programming are the separation of interface from implementation, and the notion that an object may have state. This paper describes a simple system that enables object-oriented programs to be evolved. The system exploits reflection to automatically discover features about the environment (the existing classes and objects) in which it is to operate. This enables us to evolve object-oriented programs for the given problem domain with the minimum of effort. Currently, we are only evolving method implementations. Future work will explore how we can also evolve interfaces and classes, which should be beneficial to the automatic generation of structured solutions to complex problems. We demonstrate the system with the aid of an evolutionary art example.

Evolutionary Feature Construction using Information Gain and Gini Index
Mohammed A. Muharram, George D. Smith

Abstract:
Feature construction using genetic programming is carried out to study the effect on the performance of a range of classification algorithms with the inclusion of the evolved attributes. Two different fitness functions are used in the genetic program, one based on information gain and the other based on the gini index. The classification algorithms used are three classification tree algorithms, namely C5, CART, CHAID and an MLP neural network. The intention of the research is to ascertain if the decision tree classification algorithms benefit more using features constructed using a genetic programme whose fitness function incorporates the same fundamental learning mechanism as the splitting criteria of the associated decision tree.

On the Evolution of Evolutionary Algorithms
Jorge Tavares, Penousal Machado, Amílcar Cardoso, Francisco B. Pereira, Ernesto Costa

Abstract:
In this paper we discuss the evolution of several components of a traditional Evolutionary Algorithm, such as genotype to phenotype mappings and genetic operators, presenting a formalized description of how this can be attained. We then focus on the evolution of mapping functions, for which we present experimental results achieved with a meta-evolutionary scheme.

Genetic Programming with Gradient Descent Search for Multiclass Object Classification
Mengjie Zhang, Will Smart

Abstract:
This paper describes an approach to the use of gradient descent search in genetic programming (GP) for object classification problems. Gradient descent search is introduced to the GP mechanism and is embedded into the genetic beam search, which allows the evolutionary learning process to globally follow the beam search and locally follow the gradient descent search. Two different methods, an online gradient descent scheme and an offline gradient descent scheme, are developed and compared with the basic GP method on three image data sets with object classification problems of increasing difficulty. The results suggest that both the online and the offline gradient descent GP methods outperform the basic GP method in terms of both classification accuracy and training efficiency and that the online scheme achieved better performance than the offline scheme.

Programme Co-chair: Una-May O'Reilly, MIT <unamay@ai.mit.edu>
Programme Co-chair: Maarten Keijzer, Free University Amsterdam <mkeijzer@cs.vu.nl>
Publication Chair: Terence Soule, University of Idaho <tsoule@cs.uidaho.edu>
Publicity Chair: Simon Lucas, University of Essex <sml@essex.ac.uk>
Local Chair: Ernesto Costa, University of Coimbra <ernesto@dei.uc.pt>

Programme Committee:
Vladan Babovic, Tectrasys AG
Wolfgang Banzhaf, Memorial University of Newfoundland
Bertrand Braunschweig, Institut Français du Pétrole
Martin C Martin, MIT
Stefano Cagnoni, University of Parma
Jean-Jacques Chabrier, University of Burgondy
Pierre Colet, Laboratoire d'Informatique du Littoral
Ernesto Costa, University of Coimbra
Marco Dorigo, Universite' Libre de Bruxelles
Malachy Eaton, University of Limerick
Marc Ebner, Universitaet Wuerzburg
Jeroen Eggermont, Leiden University
Aniko Ekart, Hungarian Academy of Sciences
Daryl Essam, University of New South Wales
Francisco Fernandez de Vega, University of Extremadura
Cyril Fonlupt, Université du Littoral
Alex Freitas, University of Kent
Wolfgang Golubski, University of Siegen
Steven Gustafson, University of Nottingham
Jin-Kao Hao, Universite d'Angers
Daniel Howard, QinetiQ, dhoward@qinetiq.com
Christian Jacob, University of Calgery
Colin Johnson, University of Kent at Canterbury
Didier Keymeulen, Jet Propulsion Laboratory
Bill Langdon, University College London
Simon Lucas, University of Essex
Evelyne Lutton, I.N.R.I.A. Rocquencourt
Penousal Machado, University of Coimbra
Peter Martin, Naiad Consulting Ltd.
Julian Miller, University of York
Miguel Nicolau, University of Limerick
Michael O'Neill, University of Limerick
Francisco Pereira, Instituto Superior de Engenharia de Coimbra
Riccardo Poli, University of Essex
Conor Ryan, University of Limerick
Bart Rylander, University of Portland
Kazuhiro Saitou, University of Michigan
Marc Schoenauer, I.N.R.I.A. Rocquencourt
Alexei Skourikhine, Los Alamos National Laboratory
Adrian Stoica, Jet Propulsion Laboratory
Matthew Streeter, Carnegie Mellon University
Adrian Thompson, University of Sussex
Marco Tomassini, University of Lausanne
Krister Wolff, Chalmers University of Technology
Edwin de Jong, University of Utrecht, dejong@cs.uu.nl

High quality papers are sought on topics strongly related to genetic programming, ranging from theoretical work to innovative applications.The standard is high with about 40% of submissions accepted for oral presentation, and reviewing is double blind.

Topics include:

• Theoretical developments
• Empirical studies of GP performance and behaviour
• New algorithms, representations and operators
• Applications of GP to real-life problems
• Hybrid architectures including GP components
• Comparisons with other machine learning or program-induction techniques
• New libraries and implementations
• Linear GP
• Evolution of other tree or graph structures (e.g. VRML)
• Evolution of various classes of machine: e.g. cellular automata, finite state machines, pushdown automata, turing machines
• Object-oriented genetic programming

Submissions should be a maximum of ten A4 pages and they should be sent in zipped postscript format. Papers must conform to the Springer Lecture Notes in Computer Science format: http://www.springer.de/comp/lncs/authors.html.

The reviewing process is double blind. Authors should remove their names from submitted papers, and should take reasonable care that their identity is disguised. References to own work can be included in the paper, but should be referred to in the third person.

It is very important that the email accompanying submission should state ALL the authors, including ALL their email addresses. To avoid problems with electronic delivery, papers should be emailed to BOTH of the program chairs. A notification of receipt will be emailed within three working days after the deadline.

The EuroGP2004 proceedings will be
published by Spinger as part of their
Lecture Notes in Computer Science series.