EvoCOMPLEX Logo

Detailed Programme

Evolutionary Algorithms and Complex Systems

Complex systems are ubiquitous in physics, economics, sociology, biology, computer science, and many other scientific areas. Typically, a complex system is composed of smaller aggregated components, whose interaction and interconnectedness are non-trivial (e.g., interactions can be high-dimensional and non-linear, and/or the connectivity can exhibit non-trivial topological features such as power-law degree distribution, and high clustering coefficient). This leads to emergent properties of the system, not anticipated by its isolated components. Furthermore, when the system behaviour is studied form a temporal perspective, self-organisation patterns typically arise.

Studying complex systems requires composite strategies that employ various different algorithms to solve a single difficult problem. Components of such strategies may solve consecutive phases leading to the main goal (for example, consider an oil deposit exploration strategy composed of a complex memetic search algorithm and of a direct FEM solver), may be used to approach particular sub-tasks from different perspectives (as, for example, in multi-scale approaches), or may solve the main problem in different ways that are aggregated to form the final solution (as, for example, in hyper-heuristics, island GAs or multi-physics approaches).

EvoCOMPLEX 2014 covers all aspects of the interaction of evolutionary algorithms -and metaheuristics in general- with complex systems. Topics of interest include, but are not limited to, the use of evolutionary algorithms for the analysis or design of complex systems, such as for example:

Relevant topics also include the use of complex systems and tools thereof to model, analyse or improve the performance of straightforward and complex evolutionary-based strategies evolutionary algorithms, such as for example:

PUBLICATION DETAILS

Accepted papers will appear in the proceedings of EvoStar, published in a volume of the Springer Lecture Notes in Computer Science, which will be available at the Conference.Submissions must be original and not published elsewhere. The submissions will be peer reviewed by at least three members of the program committee. 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 and attend the conference and present the work.The reviewing process will be double-blind, please omit information about the authors in the submitted paper.

Submission Details

Submissions must be original and not published elsewhere. They will be peer reviewed by 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.

Please provide up to five keywords in your Abstract

Page limit: 12 pages to http://myreview.csregistry.org/evoapps14/.

IMPORTANT DATES

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

FURTHER INFORMATION

Further information on the conference and co-located events can be
found in: http://www.evostar.org

EvoCOMPLEX Programme

Fri 1000-1140  EvoCOMPLEX 
Chair:  Carlos Cotta

Common Developmental Genomes Revisited - Evolution through Adaptation     Konstantinos Antonakopoulos
Artificial development has been widely used for designing complex structures and as a means to increase the complexity of an artifact. One central challenge in artificial development is to understand how a mapping process could work on a class of architectures in a more general way by exploiting the most favorable properties from each computational architecture or by combining efficiently more than one computational architectures (i.e., a true multicellular approach). Computational architectures in this context comprise structures with connected computational elements, namely, cellular automata and boolean networks. The ability to develop and co-evolve different computational architectures has previously been investigated using common developmental genomes. In this paper, we extend a previous work that studied their evolvability. Here, we focus on their ability to evolve when the goal changes over evolutionary time (i.e., adaptation), utilizing a more fair fitness assignment scheme. In addition, we try to investigate how common developmental genomes exploit the underlying architecture in order to build the phenotypes. The results show that they are able to find very good solutions with rather simplified solutions than anticipated.

Investigation of Genome Parameters and Sub-Transitions to Guide Evolution of Artificial Cellular Organisms     Stefano Nichele, Håkon Hjelde Wold, Gunnar Tufte
Artificial multi-cellular organisms develop from a single zygote to complex morphologies, following the instructions encoded in their genomes. Small genome mutations can result in very different developed phenotypes. In this paper we investigate how to exploit genotype information in order to guide evolution towards favorable areas of the phenotype solution space, where the sought emergent behavior is more likely to be found. Lambda genome parameter, with its ability to discriminate different developmental behaviors, is incorporated into the fitness function and used as a discriminating factor for genetic distance, to keep resulting phenotype’s developmental behavior close by and encourage beneficial mutations that yield adaptive evolution. Genome activation patterns are detected and grouped into genome parameter sub-transitions. Different sub-transitions are investigated as simple genome parameters, or composed to integrate several genome properties into a more exhaustive composite parameter. The experimental model used herein is based on 2-dimensional cellular automata.

Training Complex Decision Support Systems with Differential Evolution Enhanced by Locally Linear Embedding     Piotr Lipinski
This paper aims at improving the training process of complex decision support systems, where evolutionary algorithms are used to integrate a large number of decision rules in a form of a weighted average. It proposes an enhancement of Differential Evolution by Locally Linear Embedding to process objective functions with correlated variables, which focuses on detecting local dependencies among variables of the objective function by analyzing the manifold in the search space that contains the current population and transforming it to a reduced search space. Experiments performed on some popular benchmark functions as well as on a financial decision support system confirm that the method may significantly improve the search process in the case of objective functions with a large number of variables, which usually occur in many practical applications.

A Memetic Framework for Solving Difficult Inverse Problems     Maciej Smolka, Robert Schaefer
The paper introduces a multi-deme, memetic global optimization strategy Hierarchic memetic Strategy (HMS) especially well-suited to the solution of a class of parametric inverse problems. This strategy develops dynamically a tree of dependent populations (demes) searching with the various accuracy growing from the root to the leaves. The search accuracy is associated with the accuracy of solving direct problems by hp–adaptive Finite Element Method. Throughout the paper we describe details of exploited accuracy adaptation and computational cost reduction mechanisms, an agent-based architecture of the proposed system, a sample implementation and preliminary benchmark results.