EuroGP Logo

Detailed Programme

17th European Conference on Genetic Programming

EuroGP is the premier annual conference on Genetic Programming, the oldest and the only meeting worldwide devoted specifically to this branch of evolutionary computation. It is always a very enjoyable event attracting participants from all continents, offering excellent
opportunities for networking, informal contact, exchange of ideas and discussions with fellow researchers, in a friendly and relaxed environment. High quality papers describing new original research are sought on topics strongly related to the evolution of computer programs, ranging from theoretical work to innovative applications. The conference will feature a mixture of oral presentations and poster sessions. In 2013, the EuroGP acceptance rate was 49% (38% for oral presentations).

Areas of Interest and Contributions

Topics include but are not limited to:

Publication Details

Accepted papers will be presented orally or as posters at the Conference and will be printed in the proceedings published by Springer Verlag in the Lecture Notes in Computer Science (LNCS) series.

Previous editions of EuroGP were published in the following Springer Verlag LNCS volumes

 

 

The papers which receive the best reviews will be nominated for the Best Paper Award.

Submission Details

Submissions must be original and not published elsewhere. They will be peer reviewed by at least three members of the program committee. The reviewing process will be double-blind, so please omit information about the authors in the submitted paper. Submit your manuscript in Springer LNCS format.

Page limit: 12 pages

The authors of accepted papers will have to improve their paper on the basis of the reviewers' comments and will be asked to send a camera ready version of their manuscripts. At least one author of each accepted work has to register for the conference, attend the conference and present the work.

Submission link: http://myreview.csregistry.org/eurogp14

Important Dates

Submission deadline: 1 November 2013 11 November 2013
Notification: 06 January 2014
Camera ready: 01 February 2014
Conference dates: 23-25 April 2014

Programme Committee

Further Information

A comprehensive bibliography of genetic programming literature and links to related material is accessible at the Genetic Programming Bibliography web page, part of the Collection of Computer Science Bibliographies maintained and managed by William Langdon, Steven Gustafson, and John Koza.

Genetic Programming Algorithms

The current version of our CfP flyer is available here

EuroGP Programme

Wednesday 23 April

Wed 1120-1300 Session 1: Applications  
Chair: Atif Azad 

            Enhancing Branch-and-Bound Algorithms for Order Acceptance and Scheduling with Genetic Programming  Su Nguyen, Mengjie Zhang, Mark Johnston
Order acceptance and scheduling (OAS) is an important planning activity in make-to-order manufacturing systems. Making good acceptance and scheduling decisions allows the systems to utilise their manufacturing resources better and achieve higher total profit. Therefore, finding optimal solutions for OAS is desirable. Unfortunately, the exact optimisation approaches previously proposed for OAS are still very time consuming and usually fail to solve the problem even for small instances in a reasonable computational time. In this paper, we develop a new branch-and-bound (B&B) approach to finding optimal solutions  for OAS. In order to design effective branching strategies for B&B, a new GP method has been proposed to discover good ordering rules. The results show that the B&B algorithms enhanced by GP can solve the OAS problem more effectively than the basic B&B algorithm and the CPLEX solver on the Mixed Integer Linear Programming model.

            Building a Stage 1 Computer Aided Detector for Breast Cancer using Genetic Programming  Conor Ryan, Krzysztof Krawiec, Una-May O'Reilly, Jeannie Fitzgerald, David Medernach
We describe a fully automated workflow for performing stage1 breast cancer detection with GP as its cornerstone. Mammograms are by far the most widely used method for detecting breast cancer in women, and its use in national screening can have a dramatic impact on early detection and survival rates. With the increased availability of digital mammography, it is becoming increasingly more feasible to use auto- mated methods to help with detection. A stage 1 detector examines mammograms and highlights suspicious areas that require further investigation. A too conservative approach degenerates to marking every mammogram (or segment of) as suspicious, while missing a cancerous area can be disastrous. Our workflow positions us right at the data collection phase such that we generate textural features ourselves. These are fed through our system, which performs PCA on them before passing the most salient ones to GP to generate classifiers. The classifiers give results of 100% accuracy on true positives and a false positive per image rating of just 1.5, which is better than prior work. Not only this, but our system can use GP as part of a feedback loop, to both select and help generate further features.

            Exploring the Search Space of Hardware/Software Embedded Systems by Means of GP  Milos Minarik, Lukas Sekanina
This paper presents a new platform for development of small application-specific digital embedded architectures based on a data path controlled by a microprogram. Linear genetic programming is extended to evolve a program for the controller together with suitable hardware architecture. Experimental results show that the platform can automatically design general solutions as well as highly optimized specialized solutions to benchmark problems such as maximum, parity or iterative division.

Wed 1430-1610 Session 2: Operators 
Chair: Sara Silva

            Semantic Crossover based on the Partial Derivative Error  Mario Graff, Ariel Graff-Guerrero, Jaime Cerda-Jacobo
There is great interest for the development of semantic genetic operators to improve the performance of genetic programming. Semantic genetic operators have traditionally been developed employing experimentally or theoretically-based approaches. Our current work proposes a novel semantic crossover developed amid the two traditional approaches. Our proposed semantic crossover operator is based on the use of the derivative of the error propagated through the tree. This process decides the crossing point of the second parent. The results show that our procedure improves the performance of genetic programming on rational symbolic regression problems.

            Measuring Mutation Operators' Exploration-Exploitation Behaviour and Long-term Biases  James McDermott
We propose a simple method of directly measuring a mutation operator's short-term exploration-exploitation behaviour, based on its transition matrix. Higher values for this measure indicate a more exploitative operator. Since operators also differ in their degree of long-term bias towards particular areas of the search space, we propose a simple method of directly measuring this bias, based on the Markov chain stationary state. We use these measures to compare numerically the behaviours of two well-known mutation operators, the genetic algorithm per-gene bitflip mutation and the genetic programming subtree mutation.

            NEAT, There's no Bloat  Leonardo Trujillo, Luis Muñoz, Enrique Naredo, Yuliana Martinez
The Operator Equalization (OE) family of bloat control methods have achieved promising results in many domains. In particular, the Flat-OE method, that promotes a flat distribution of program sizes, is one of the simplest OE methods and achieves some of the best results. However, Flat-OE, like all OE variants, can be computationally expensive. This work proposes a simplified strategy for bloat control based on Flat-OE. In particular, bloat is studied in the NeuroEvolution of Augmenting Topologies (NEAT) algorithm. NEAT includes a very simple diversity preservation technique based on speciation and fitness sharing, and it is hypothesized that with some minor tuning, speciation in NEAT can promote a flat distribution of program size. Results indicate that this is the case in two benchmark problems, in accordance with results for Flat-OE. In conclusion, NEAT provides a worthwhile strategy that could be extrapolated to other GP systems, for effective and simple bloat control.

Wed 1745-1900  EuroGP posters

Asynchronous Evolution by Reference-based Evaluation: Tertiary Parent Selection and its Archive  Tomohiro Harada, Keiki Takadama
This paper proposes a novel asynchronous reference-based evaluation (named as ARE) for an asynchronous EA that evolves individuals independently unlike general EAs that evolve all individuals at the same time. ARE is designed for an asynchronous evolution by tertiary parent selection and its archive. In particular, ARE asynchronously evolves individuals through a comparison with only three of individuals (i.e., two parents and one reference individual as the tertiary parent). In addition, ARE builds an archive of good reference individuals. This differ from synchronous evolution in EAs in which selection involves comparison with all population members. In this paper, we investigate the effectiveness of ARE, by applying it to some standard problems used in Linear GP that aim being to minimize the execution step of machine-code programs. We compare GP using ARE (ARE-GP) with steady state (synchronous) GP (SSGP) and our previous asynchronous GP (Tierra-based Asynchronous GP: TAGP). The experimental results have revealed that ARE-GP not only asynchronously evolves the machine-code programs, but also outperforms SSGP and TAGP in all test problems.

Behavioral Search Drivers for Genetic Programing  Krzysztof Krawiec, Una-May O'Reilly
Synthesizing a program with the desired input-output behavior by means of genetic programming is an iterative process that needs appropriate guidance. That guidance is conventionally provided by a fitness function that measures the conformance of program output with the desired output. Contrary to widely adopted stance, there is no evidence that this quality measure is the best choice; alternative search drivers may exist that make search more effective. This study proposes and investigates a new family of behavioral search drivers, which inspect not only final program output, but also program behavior meant as the partial results it arrives at while executed.

Cartesian Genetic Programming: Why No Bloat?  Andrew Turner, Julian Miller
For many years now it has been known that Cartesian Genetic Programming (CGP) does not exhibit program bloat. Two possible explanations have been proposed in the literature: neutral genetic drift and length bias. This paper empirically disproves both of these and thus, reopens the question as to why CGP does not suffer from bloat. It has also been shown for CGP that using a very large number of nodes considerably increases the effectiveness of the search. This paper also proposes a new explanation as to why this may be the case.

 

On Evolution of Multi-Category Pattern Classifiers Suitable for Embedded Systems
Zdenek Vasicek, Michal Bidlo
This paper addresses the problem of evolutionary design of classifiers for the recognition of handwritten digit symbols by means of Cartesian Genetic Programming.  Two different design scenarios are investigated - the design of multiple-output classifier, and design of multiple binary classifiers.  The goal is to evolve classification algorithms that employ substantially smaller amount of operations in contrast with conventional approaches such as Support Vector Machines.  Even if the evolved classifiers do not reach the accuracy of the tuned SVM classifier, it will be shown that the accuracy is higher than 93% and the number of required operations is a magnitude lower.

 

The Best Things Don't Always Come in Small Packages: Constant Creation in Grammatical Evolution  R. Muhammad Atif Azad, Conor Ryan
This paper evaluates the performance of various methods to constant creation in
Grammatical Evolution (GE), and validates the results against those from Genetic
Programming (GP). Constant creation in GE is an important issue due to the disruptive nature of ripple crossover, which can radically remap multiple terminals in an individual, and we investigate if more compact methods, which are more similar to the GP style of constant creation (Ephemeral Random Constants (ERCs), perform better.  The results are surprising. The GE methods all perform significantly better than GP on unseen test data, and we demonstrate that the standard GE approach of digitconcatenation does not produce individuals that are any larger than those from methods which are designed to use less genetic material.

Thursday 24 April

Thurs 0930-1110 Session 3:  Software Engineering and Software Frameworks
Chair: Evelyne Lutton

            Genetically Improved CUDA C++ Software  William B. Langdon, Mark Harman
Genetic Programming (GP) may dramatically increase the performance of software written by domain experts. GP and autotuning are used to optimise and refactor legacy GPGPU C code for modern parallel graphics hardware and software. Speed ups of more than six times on recent nVidia GPU cards are reported compared to the original kernel on the same hardware.

            Using Genetic Improvement and Code Transplants to Specialise a C++ Program to a Problem Class  Justyna Petke, Mark Harman, William B. Langdon, Westley Weimer
Genetic Improvement (GI) is a form of Genetic Programming that improves an existing program. We use GI to evolve a faster version of a  C++ program, a Boolean satisfiability (SAT) solver called MiniSAT, specialising it for a particular problem class, namely Combinatorial Interaction Testing (CIT), using automated code transplantation. Our GI-evolved solver achieves overall 17% improvement, making it comparable with average expert human performance. Additionally, this automatically evolved solver is faster than any of the human-improved solvers for the CIT problem.

            Flash: A GP-GPU Ensemble Learning System for handling Large Datasets  Ignacio Arnaldo, Kalyan Veeramachaneni, Una-May O'Reilly
The Flash system runs ensemble-based Genetic Programming (GP) symbolic regression on a shared memory desktop.  To significantly reduce the high time cost of the extensive model predictions required by symbolic regression, its fitness evaluations are tasked to the desktop's GPU. Successive GP "instances" are run on different data subsets and randomly chosen objective functions. Best models are collected after a fixed number of generations and then fused with an adaptive, output-space method. New instance launches are halted once learning is complete. We demonstrate that Flash's ensemble strategy not only makes GP more robust, but it also provides an informed online means of halting the learning process. Flash enables GP to learn from a dataset composed of 370K exemplars and 90 features, evolving a population of 1000 individuals over 100 generations in as few as 50 seconds.

Thurs 1135-1315  Session 4: Theory and Analysis 
Chair: Mengjie Zhang

Higher Order Functions for Kernel Regression  Alexandros Agapitos, James McDermott, Michael O'Neill, Ahmed Kattan, Anthony Brabazon
Kernel regression is a well-established nonparametric method, in which the target value of a query point is estimated using a weighted average of the surrounding training examples. The weights are typically obtained by applying a distance-based kernel function, which presupposes the existence of a distance measure. This paper investigates the use of Genetic Programming for the evolution of task-specific distance measures as an alternative to Euclidean distance. Results on seven real-world datasets show that the generalisation performance of the proposed system is superior to that of Euclidean-based kernel regression and standard GP.

            A Multi-dimensional Genetic Programming Approach for Multi-class Classification Problems  Vijay Ingalalli, Sara Silva, Mauro Castelli, Leonardo Vanneschi
Classification problems are of profound interest for the machine learning community as well as to an array of application fields. However, multi-class classification problems can be very complex, in particular when the number of classes is high. Although very successful in so many applications, GP was never regarded as a good method to perform multi-class classification. In this work, we present a novel algorithm for tree based GP, that incorporates some ideas on the representation of the solution space in higher dimensions. This idea lays some foundations on addressing multi-class classification problems using GP, which may lead to further research in this direction. We test the new approach on a large set of benchmark problems from several different sources, and observe its competitiveness against the most successful state-of-the-art classifiers.

            On Diversity, Teaming and Hierarchical Policies: Observations from the Keepaway Soccer Task  Stephen Kelly, Malcolm Heywood
The 3-versus-2 Keepaway soccer task represents a widely used benchmark appropriate for evaluating approaches to reinforcement learning, multi-agent systems, and evolutionary robotics. To date most research on this task has been described in terms of developments to reinforcement learning with function approximation or frameworks for neuro-evolution. This work performs an initial study using a recently proposed algorithm for evolving teams of programs hierarchically using two phases of evolution: one to build a library of candidate meta policies and a second to learn how to deploy the library consistently. Particular attention is paid to diversity maintenance, where this has been demonstrated as a critical component in neuro-evolutionary approaches. A new formulation is proposed for fitness sharing appropriate to the Keepaway task. The resulting policies are observed to benefit from the use of diversity and perform significantly better than previously reported. Moreover, champion individuals evolved and selected under one field size generalize to multiple field sizes without any additional training.

 

 

Thurs 1430-1610 Session 5: EuroGP best paper nominations 
Chairs: Miguel Nicolau and Krzysztof Krawiec

            ESAGP - A Semantic GP Framework Based on Alignment in the Error Space  (EuroGP best paper candidate)  Stefano Ruberto, Leonardo Vanneschi, Mauro Castelli, Sara Silva
This paper introduces the concepts of error vector and error space, directly bound to semantics, one of the hottest topics in genetic programming. Based on these concepts, we introduce the notions of optimally aligned individuals and optimally coplanar individuals. We show that, given optimally aligned, or optimally coplanar, individuals, it is possible to construct a globally optimal solution analytically. Thus, we introduce a genetic programming framework for symbolic regression called Error Space Alignment GP (ESAGP) and two of its instances: ESAGP-1, whose objective is to find optimally aligned individuals, and ESAGP-2, whose objective is to find optimally coplanar individuals. We also discuss how to generalize the approach to any number of dimensions. Using two complex real-life applications, we provide experimental evidence that ESAGP-2 outperforms ESAGP-1, which in turn outperforms both standard GP and geometric semantic GP. This suggests that ``adding dimensions'' is beneficial and encourages us to pursue the study in many different directions, that we summarize in the final part of the manuscript.

            Generalisation Enhancement via Input Space Transformation: A GP Approach  (EuroGP best paper candidate)  Ahmed Kattan, Michael Kampouridis, Alexandros Agapitos
This paper proposes a new approach to improve generalisation of standard regression techniques when there are hundreds or thousands of input variables. The input space $X$ is composed of observational data of the form $(x_i, y(x_i)), i = 1... n$ where each $x_i$ denotes a k-dimensional input vector of design variables and $y$ is the response. Genetic Programming (GP) is used to transform the original input space $X$ into a new input space $Z = (z_i, y(z_i))$ that has smaller input vector and is easier to be mapped into its corresponding responses.  GP is designed to evolve a function that receives the original input vector from each $x_i$ in the original input space  as input and return a new vector $z_i$ as an output. Each element in the newly evolved $z_i$ vector is generated from an evolved mathematical formula that extracts statistical features from the original input space. To achieve this, we designed GP trees to produce multiple outputs. Empirical evaluation of $20$ different problems revealed that the new approach is able to significantly reduce the dimensionality of the original input space and improve the performance of standard approximation models such as Kriging, Radial Basis Functions Networks, and Linear Regression, and GP (as a regression techniques). In addition, results demonstrate that the new approach is better than standard dimensionality reduction techniques such as Principle Component Analysis (PCA). Moreover, the results show that the proposed approach is able to improve the performance of standard Linear Regression and make it competitive to other stochastic regression techniques.

            Learning Dynamical Systems Using Standard Symbolic Regression  (EuroGP best paper candidate)  Sébastien Gaucel, Maarten Keijzer, Evelyne Lutton, Alberto Tonda

Symbolic regression has many successful applications in learning free-form regular equations from data. Trying to apply the same approach to differential equations is the logical next step: so far, however, results have not matched the quality obtained with regular equations, mainly due to additional constraints and dependencies between variables that make the problem extremely hard to tackle. In this paper we propose a new approach to dynamic systems learning. Symbolic regression is used to obtain a set of first-order Eulerian approximations of differential equations, and mathematical properties of the approximation are then exploited to reconstruct the original differential equations. Advantages of this technique include the de-coupling of systems of differential equations, that can now be learned independently; the possibility of exploiting established techniques for standard symbolic regression, after trivial operations on the original dataset; and the substantial reduction of computational effort, when compared to existing ad-hoc solutions for the same purpose. Experimental results show the efficacy of the proposed approach on an instance of the Lotka-Volterra model.

Thurs 1630-1810  Session 5: EuroGP Debate
Chairs: Miguel Nicolau and Krzysztof Krawiec

EuroGP invites all to an open session to discuss issues and current trends in the field of Genetic Programming.