ECTA 2015 Abstracts


Full Papers
Paper Nr: 2
Title:

A Heuristic Solution for Noisy Image Segmentation using Particle Swarm Optimization and Fuzzy Clustering

Authors:

Saeed Mirghasemi, Ramesh Rayudu and Mengjie Zhang

Abstract: Introducing methods that can work out the problem of noisy image segmentation is necessary for real-world vision problems. This paper proposes a new computational algorithm for segmentation of gray images contaminated with impulse noise. We have used Fuzzy C-Means (FCM) in fusion with Particle Swarm Optimization (PSO) to define a new similarity metric based on combining different intensity-based neighborhood features. PSO as a computational search algorithm, looks for an optimum similarity metric, and FCM as a clustering technique, helps to verify the similarity metric goodness. The proposed method has no parameters to tune, and works adaptively to eliminate impulsive noise. We have tested our algorithm on different synthetic and real images, and provided quantitative evaluation to measure effectiveness. The results show that, the method has promising performance in comparison with other existing methods in cases where images have been corrupted with a high density noise.

Paper Nr: 3
Title:

On Routine Evolution of New Replicating Structures in Cellular Automata

Authors:

Michal Bidlo

Abstract: This paper presents evolutionary design of two-dimensional, uniform cellular automata. The problem of replicating loops is considered as a case study. Conditionally matching rules are used as a technique that is suitable to the design of cellular automata state transition rules. A genetic algorithm is applied to the design of cellular automata that satisfy the requirements of replicating loops. It is shown that such evolution is able to find various state transition rules that support replication of a given loop. Results presented herein demonstrate the ability of derived cellular automata to perform replication not only from an initial instance of the loop but also, that from a seed the loop can autonomously grow.

Paper Nr: 5
Title:

Clustering Analysis using Opposition-based API Algorithm

Authors:

Mohammad Reza Farmani and Giuliano Armano

Abstract: Clustering is a significant data mining task which partitions datasets based on similarities among data. In this study, partitional clustering is considered as an optimization problem and an improved ant-based algorithm, named Opposition-Based API (after the name of Pachycondyla APIcalis ants), is applied to automatic grouping of large unlabeled datasets. The proposed algorithm employs Opposition-Based Learning (OBL) for ants' hunting sites generation phase in API. Experimental results are compared with the classical API clustering algorithm and three other recently evolutionary-based clustering techniques. It is shown that the proposed algorithm can achieve the optimal number of clusters and, in most cases, outperforms the other methods on several benchmark datasets in terms of accuracy and convergence speed.

Paper Nr: 13
Title:

Application of Adaptive Differential Evolution for Model Identification in Furnace Optimized Control System

Authors:

Miguel Leon, Magnus Evestedt and Ning Xiong

Abstract: Accurate system modelling is an important prerequisite for optimized process control in modern industrial scenarios. The task of parameter identification for a model can be considered as an optimization problem of searching for a set of continuous parameters to minimize the discrepancy between the model outputs and true output values. Differential Evolution (DE), as a class of population-based and global search algorithms, has strong potential to be employed here to solve this problem. Nevertheless, the performance of DE is rather sensitive to its two running parameters: scaling factor and crossover rate. Improper setting of these two parameters may cause weak performance of DE in real applications. This paper presents a new adaptive algorithm for DE, which does not require good parameter values to be specified by users in advance. Our new algorithm is established by integration of greedy search into the original DE algorithm. Greedy search is conducted repeatedly during the running of DE to reach better parameter assignments in the neighborhood. We have applied our adaptive DE algorithm for process model identification in a Furnace Optimized Control System (FOCS). The experiment results revealed that our adaptive DE algorithm yielded process models that estimated temperatures inside a furnace more precisely than those produced by using the original DE algorithm.

Paper Nr: 16
Title:

Multi-Strategy Genetic Algorithm for Multimodal Optimization

Authors:

Evgenii Sopov

Abstract: Multimodal optimization (MMO) is the problem of finding many or all global and local optima. In recent years many efficient nature-inspired techniques (based on ES, PSO, DE and others) have been proposed for real-valued problems. Many real-world problems contain variables of many different types, including integer, rank, binary and others. In this case, the weakest representation (namely binary representation) is used. Unfortunately, there is a lack of efficient approaches for problems with binary representation. Existing techniques are usually based on general ideas of niching. Moreover, there exists the problem of choosing a suitable algorithm and fine tuning it for a certain problem. In this study, a novel approach based on a metaheuristic for designing multi-strategy genetic algorithm is proposed. The approach controls the interactions of many search techniques (different genetic algorithms for MMO) and leads to the self-configuring solving of problems with a priori unknown structure. The results of numerical experiments for classical benchmark problems and benchmark problems from the CEC competition on MMO are presented. The proposed approach has demonstrated efficiency better than standard niching techniques and comparable to advanced algorithms. The main feature of the approach is that it does not require the participation of the human-expert, because it operates in an automated, self-configuring way.

Paper Nr: 25
Title:

GPGPU Vs Multiprocessor SPSO Implementations to Solve Electromagnetic Optimization Problems

Authors:

Anton Duca, Laurentiu Duca, Gabriela Ciuprina and Daniel Ioan

Abstract: This paper studies two parallelization techniques for the implementation of a SPSO algorithm applied to optimize electromagnetic field devices, GPGPU and Pthreads for multiprocessor architectures. The GPGPU and Pthreads implementations are compared in terms of solution quality and speed up. The electromagnetic optimization problems chosen for testing the efficiency of the parallelization techniques are the TEAM22 benchmark problem and Loney’s solenoid problem. As we will show, there is no single best parallel implementation strategy since the performances depend on the optimization function.

Paper Nr: 26
Title:

Multi-Robot Cooperative Tasks using Combined Nature-Inspired Techniques

Authors:

Nunzia Palmieri, Floriano de Rango, Xin She Yang and Salvatore Marano

Abstract: In this paper, two metaheuristics are presented for exploration and mine disarming tasks performed by a swarm of robots. The objective is to explore autonomously an unknown area in order to discover the mines, disseminated in the area, and disarm them in cooperative manner since a mine needs multiple robots to disarm. The problem is bi-objective: distributing in different regions the robots in order to explore the area in a minimum amount of time and recruiting the robots in the same location to disarm the mines. While autonomous exploration has been investigated in the past, we specifically focus on the issue of how the swarm can inform its members about the detected mines, and guide robots to the locations. We propose two bio-inspired strategies to coordinate the swarm: the first is based on the Ant Colony Optimization (ATS-RR) and the other is based on the Firefly Algorithm (FTS-RR). Our experiments were conducted by simulations evaluating the performance in terms of exploring and disarming time and the number of accesses in the operative grid area applying both strategies in comparison with the Particle Swarm Optimization (PSO). The results show that FTS-RR strategy performs better especially when the complexity of the tasks increases.

Paper Nr: 31
Title:

GEML: Evolutionary Unsupervised and Semi-Supervised Learning of Multi-class Classification with Grammatical Evolution

Authors:

Jeannie M. Fitzgerald, R. Mohammed Atif Azad and Conor Ryan

Abstract: This paper introduces a novel evolutionary approach which can be applied to supervised, semi-supervised and unsupervised learning tasks. The method, Grammatical Evolution Machine Learning (GEML) adapts machine learning concepts from decision tree learning and clustering methods, and integrates these into a Grammatical Evolution framework. With minor adaptations to the objective function the system can be trivially modified to work with the conceptually different paradigms of supervised, semi-supervised and unsupervised learning. The framework generates human readable solutions which explain the mechanics behind the classification decisions, offering a significant advantage over existing paradigms for unsupervised and semi-supervised learning. GEML is studied on a range of multi-class classification problems and is shown to be competitive with several state of the art multi-class classification algorithms.

Paper Nr: 32
Title:

For Sale or Wanted: Directed Crossover in Adjudicated Space

Authors:

Jeannie M. Fitzgerald and Conor Ryan

Abstract: Significant recent effort in genetic programming has focused on selecting and combining candidate solutions according to a notion of behaviour defined in semantic space and has also highlighted disadvantages of relying on a single scalar measure to capture the complexity of program performance in evolutionary search. In this paper, we take an alternative, yet complementary approach which directs crossover in what we call adjudicated space, where adjudicated space represents an abstraction of program behaviour that focuses on the success or failure of candidate solutions in solving problem sub-components. We investigate the effectiveness of several possible adjudicated strategies on a variety of classification and symbolic regression problems, and show that both of our novel pillage and barter tactics significantly outperform both a standard genetic programming and an enhanced genetic programming configuration on the fourteen problems studied.

Paper Nr: 37
Title:

Design of an Autonomous Intelligent Demand-Side Management System by using Electric Vehicles as Mobile Energy Storage Units by Means of Evolutionary Algorithms

Authors:

Edgar Galván-Lopez, Marc Schoenauer and Constantinos Patsakis

Abstract: Evolutionary Algorithms (EAs), or Evolutionary Computation, are powerful algorithms that have been used in a range of challenging real-world problems. In this paper, we are interested in their applicability on a dynamic and complex problem borrowed from Demand-Side Management (DSM) systems, which is a highly popular research area within smart grids. DSM systems aim to help both end-use consumer and utility companies to reduce, for instance, peak loads by means of programs normally implemented by utility companies. In this work, we propose a novel mechanism to design an autonomous intelligent DSM by using (EV) electric vehicles’ batteries as mobile energy storage units to partially fulfill the energy demand of dozens of household units. This mechanism uses EAs to automatically search for optimal plans, representing the energy drawn from the EVs’ batteries. To test our approach, we used a dynamic scenario where we simulated the consumption of 40 and 80 household units over a period of 30 working days. The results obtained by our proposed approach are highly encouraging: it is able to use the maximum allowed energy that can be taken from each EV for each of the simulated days. Additionally, it uses the most amount of energy whenever it is needed the most (i.e., high-peak periods) resulting into reduction of peak loads.

Paper Nr: 45
Title:

A GISMOO Algorithm for a Multi-Objective Permutation Flowshop with Sequence-Dependent Setup Times

Authors:

Aymen Sioud, Caroline Gagné and julien Dort

Abstract: This paper presents a GISMOO algorithm adaptation to solve a multi-objective permutation flowshop with sequence-dependent setup times. The makespan and the total tardiness are the two objectives studied. Numerical experiments on various benchmarks from the literature were performed, to compare the performance of the adapted GISMOO algorithm with the NSGA-II algorithm.The results indicate that our algorithm generates better solutions than the known reference algorithms.

Paper Nr: 51
Title:

Particle Convergence Time in the PSO Model with Inertia Weight

Authors:

Krzysztof Trojanowski and Tomasz Kulpa

Abstract: Particle Swarm Optimization (PSO) is a powerful heuristic optimization method being subject of continuous interest. Theoretical analysis of its properties concerns primarily the conditions necessary for guaranteeing its convergent behaviour. Particle behaviour depends on three groups of parameters: values of factors in a velocity update rule, initial localization and velocity and fitness landscape. The paper presents theoretical analysis of the particle convergence properties in the model with inertia weight respectively to different values of these parameters. A new measure for evaluation of a particle convergence time is proposed. For this measure an upper bound formula is derived and its four main types of characteristics are discussed. The way of the characteristics transformations respectively to changes of velocity equation parameters is presented as well.

Short Papers
Paper Nr: 4
Title:

Supply of Order-1 Building Blocks for Functions Linearly Combined of Sinusoidal Bases with Integral Frequencies

Authors:

Hongqiang Mo, Zhong Li and Qiliang Du

Abstract: In line with the theory of schema sampling, a hypothesis could be made that sufficient supply of loworder building blocks (BBs) was one of the necessary conditions for a genetic algorithm(GA) to work. A consequential question of this hypothesis regards, when a certain fitness function is optimized with a commonly used GA, whether it is rare or common that there are plenty of low-order BBs. It is remarked that, when a base-m encoded GA is applied to a fitness function that is linearly combined of sinusoidal basis functions with integral frequencies, it is unlikely to obtain order-1 BBs with fixed positions at multiple loci, i.e., it is rare that there are plenty of order-1 BBs. However, if a considerable part of the sinusoidal basis functions are with frequencies exponential to a positive integer m, a base-m encoding can provide relatively more order-1 BBs compared with the encodings with cardinalities other than m.

Paper Nr: 12
Title:

Genetic Algorithm as Machine Learning for Profiles Recognition

Authors:

Yann Carbonne and Christelle Jacob

Abstract: Persons are often asked to provide information about themselves. These data are very heterogeneous and result in as many “profiles” as contexts. Sorting a large amount of profiles from different contexts and assigning them back to a specific individual is quite a difficult problem. Semantic processing and machine learning are key tools to achieve this goal. This paper describes a framework to address this issue by means of concepts and algorithms selected from different Artificial Intelligence fields. Indeed, a Vector Space Model is customized to first transpose semantic information into a mathematical model. Then, this model goes through a Genetic Algorithm (GA) which is used as a supervised learning algorithm for training a computer to determine how much two profiles are similar. Amongst the GAs, this study introduces a new reproduction method (Best Together), and compare it to some usual ones (Wheel, Binary Tournament).This paper also evaluates the accuracy of the GAs predictions for profiles clustering with the computation of a similarity score, as well as its ability to classify two profiles are similar or non-similar. We believe that the overall methodology can be used for any kind of sources using profiles and, more generally, for similar data recognition. 1

Paper Nr: 21
Title:

Examining the Impact of Neutral Theory on Genetic Algorithm Population Evolution

Authors:

Seamus Hill and Colm O'Riordan

Abstract: This paper examines the introduction of neutrality as proposed by Kimura (Kimura, 1968) into the genotype-phenotype mapping of a Genetic Algorithm (GA). The paper looks at the evolution of both a simple GA (SGA) and a multi-layered GA (MGA) incorporating a layered genotype-phenotype mapping based on the biological concepts of Transcription and Translation. Previous research in comparing GAs often use performance statistics; in this paper an analysis of population dynamics is used for comparison. Results illustrate that the MGA population’s evolution trajectory is quite different to that of the SGA population over dynamic landscapes and that the introduction of neutrality implicitly maintains genetic diversity within the population primarily through genetic drift in association with selection.

Paper Nr: 28
Title:

Towards Finding an Effective Way of Discrete Problems Solving: The Particle Swarm Optimization, Genetic Algorithm and Linkage Learning Techniques Hybrydization

Authors:

Bartosz Andrzej Fidrysiak and Michal Przewozniczek

Abstract: Particle Swarm Optimization (PSO) and Genetic Algorithms (GA) are well known optimization tools. PSO advantage is its capability for fast convergence to the promising solutions. On the other hand GAs are able to process schemata thanks to the use of crossover operator. However, both methods have also their drawbacks – PSO may fall into the trap of preconvergence, while GA capability of fast finding locally optimal (or close to optimal) solutions seems low when compared to PSO. Relatively new, important research direction in the field of Evolutionary Algorithms is linkage learning. The linkage learning methods gather the information about possible gene dependencies and use it to improve their effectiveness. Recently, the linkage learning evolutionary methods were shown to be effective tools to solve both: theoretical and practical problems. Therefore, this paper proposes a PSO and GA hybrid, improved by the linkage learning mechanisms, dedicated to solve binary problems. The proposed method tries to combine the GA schema processing ability, linkage information processing and uses fast PSO convergence to quickly improve the quality of already known solutions.

Paper Nr: 29
Title:

An Evolutionary and Graph-Rewriting based Approach to Graph Generation

Authors:

Aaron Barry, Josephine Griffith and Colm O'Riordan

Abstract: This paper describes an evolutionary computation based graph rewriting approach to generating classes of graphs that exhibit a set of desired global features. A set of rules are used to generate, in a constructive manner, classes of graphs. Each rule represents a transformation from one graph to another. Each of these transformations causes local changes in the graph. Probabilities can be assigned to the rules which govern the frequency with which they will be applied. By assigning these probabilities correctly, one can generate graphs exhibiting desirable global features. However, choosing the correct probability distribution to generate the desired graphs is not an easy task for certain graphs and the task of finding the correct settings for these graphs may represent a difficult search space for the evolutionary algorithms. In order to generate graphs exhibiting desirable features, an evolutionary algorithm is used to find the suitable probabilities to assign to the rules. The fitness function rewards graphs that exhibit the desired properties. We show, using a small rule base, how a range of graphs can be generated.

Paper Nr: 30
Title:

An Improved Single Node Genetic Programming for Symbolic Regression

Authors:

Jiří Kubalík and Robert Babuška

Abstract: This paper presents a first step of our research on designing an effective and efficient GP-based method for solving the symbolic regression. We have proposed three extensions of the standard Single Node GP, namely (1) a selection strategy for choosing nodes to be mutated based on the depth of the nodes, (2) operators for placing a compact version of the best tree to the beginning and to the end of the population, and (3) a local search strategy with multiple mutations applied in each iteration. All the proposed modifications have been experimentally evaluated on three symbolic regression problems and compared with standard GP and SNGP. The achieved results are promising showing the potential of the proposed modifications to significantly improve the performance of the SNGP algorithm.

Paper Nr: 35
Title:

There is Noisy Lunch: A Study of Noise in Evolutionary Optimization Problems

Authors:

Juan J. Merelo, Federico Liberatore, Antonio Fernández Ares, Rubén García, Zeineb Chelly, Carlos Cotta, Nuria Rico, Antonio M. Mora and Pablo García-Sánchez

Abstract: Noise or uncertainty appear in many optimization processes when there is not a single measure of optimality or fitness but a random variable representing it. These kind of problems have been known for a long time, but there has been no investigation of the statistical distribution those random variables follow, assuming in most cases that it is distributed normally and, thus, it can be modelled via an additive or multiplicative noise on top of a non-noisy fitness. In this paper we will look at several uncertain optimization problems that have been addressed by means of Evolutionary Algorithms and prove that there is no single statistical model the evaluations of the fitness functions follow, being different not only from one problem to the next, but in different phases of the optimization in a single problem.

Paper Nr: 36
Title:

Multi-Objective Optimization using Microgenetic Algorithm Applied to the Placement of Remote and Manual Switches in Distribution Networks

Authors:

Helton do Nascimento Alves and Railson Severiano de Sousa

Abstract: This paper presents a new formulation for placement of switches in distribution of electric power. An approach for determination of the location of tie switches and section switches using a multi-objective microgenetic algorithm is proposed. In the procedure, load importance, reliability index, remote or manual controlled switch and investments costs are considered. The results are based on simulations in a 69-bus test system presented and the results are compared to the solution given by others search techniques. This comparison confirms the efficiency of the proposed method which makes it promising to solve complex problems of tie switches and section switches placement in distribution feeders.

Paper Nr: 38
Title:

FPGA Implementation of a Multi-Population PBIL Algorithm

Authors:

João Paulo Coelho, Tatiana M. Pinho and José Boaventura-Cunha

Abstract: Evolutionary-based algorithms play an important role in finding solutions to many problems that are not solved by classical methods, and particularly so for those cases where solutions lie within extreme non-convex multidimensional spaces. The intrinsic parallel structure of evolutionary algorithms are amenable to the simultaneous testing of multiple solutions; this has proved essential to the circumvention of local optima, and such robustness comes with high computational overhead, though custom digital processor use may reduce this cost. This paper presents a new implementation of an old, and almost forgotten, evolutionary algorithm: the population-based incremental learning method. We show that the structure of this algorithm is well suited to implementation within programmable logic, as compared with contemporary genetic algorithms. Further, the inherent concurrency of our FPGA implementation facilitates the integration and testing of micro-populations.

Paper Nr: 40
Title:

GA-based Action Learning

Authors:

Satoshi Yonemoto

Abstract: This paper describes a GA-based action learning framework. First, we propose a GA-based method for action learning. In this work, GA is used to learn perception-action rules that cannot be represented as genes directly. The chromosome with the best fitness (elitist) acquires the perception-action rules through the learning process. And then, we extend the method to action series learning. In the extended method, action series can be treated as one of perception-action rules. We present the experimental results of three controllers (simple game AI testbed) using the GA-based action learning framework.

Paper Nr: 41
Title:

Investigation into Mutation Operators for Microbial Genetic Algorithm

Authors:

Samreen Umer

Abstract: Microbial Genetic Algorithm (MGA) is a simple variant of genetic algorithm and is inspired by bacterial conjugation for evolution. In this paper we have discussed and analyzed variants of this less exploited algorithm on known benchmark testing functions to suggest a suitable choice of mutation operator. We also proposed a simple adaptive scheme to adjust the impact of mutation according to the diversity in population in a cost effective way. Our investigation suggests that a clever choice of mutation operator can enhance the performance of basic MGA significantly.

Paper Nr: 46
Title:

Cybersecurity and Honeypots: Experience in a Scientific Network Infrastructure

Authors:

Juan Luis Martin Acal, Gustavo Romero López, Pablo Palacín Gómez, Pablo García Sánchez, Juan Julián Merelo Guervós and Pedro A. Castillo Valdivieso

Abstract: When dealing with security concerns in the use of network infrastructures a good balance between security concerns and the right to privacy should be maintained. This is very important in scientific networks, because they were created with an open and decentralized philosophy, in favor of the transmission of knowledge, when security was not a essential topic. Although private and scientific information have an enormous value for an attacker, the user privacy for legal and ethical reasons must be respected. Thus, passive detection methods in cybersecurity such as honeypots are a good strategy to achieve this balance between security and privacy in the defense plan of a scientific network. In this paper we present the practical case of the University of Granada in the application of honeypots for the detection and study of intrusions, which avoid intrusive techniques such as the direct analysis of the traffic through networking devices.

Paper Nr: 49
Title:

Ephemeral Computing and Bioinspired Optimization - Challenges and Opportunities

Authors:

Carlos Cotta, Antonio J. Fernández-Leiva, Francisco Fernández de Vega, Francisco Chávez, Juan J. Merelo, Pedro A. Castillo, David Camacho and Gema Bello-Orgaz

Abstract: Computational devices with significant computing power are pervasive yet often under-exploited since they are frequently idle or performing non-demanding tasks. Exploiting this power can be a cost-effective solution for solving complex computational tasks. Device-wise, this computational power can some times comprise a stable, long-lasting availability windows but it will more frequently take the form of brief, ephemeral bursts, mainly in the presence of devices “lent” voluntarily by their users. A highly dynamic and volatile computational landscape emerges from the collective contribution of numerous such devices. Algorithms consciously running on these environments require specific properties in terms of flexibility, plasticity and robustness. Bioinspired algorithms are particularly well suited to this endeavor, thanks to their intrinsic features: decentralized functioning, intrinsic parallelism, resilience, and adaptiveness. The latter is essential to exert advanced self-control on the functioning and/or structure of the algorithm. Much has been done in providing self-adaptation capabilities to these techniques, yet the science of self-? bionspired algorithms is still nascent, in particular regarding to higher-level self-adaptation, and self-management in the context of large scale optimization problems and distributed ephemeral computing technologies. Deploying bioinspired techniques on this scenario will also pave the way for the application of other techniques on this computational domain.

Paper Nr: 53
Title:

The Vantage Point Bees Algorithm

Authors:

Sultan Zeybek and Ebubekir Koç

Abstract: In this paper, an implementation of vantage point local search procedure for the Bees Algorithm (BA) in combinatorial domains is presented. In its basic version, the BA employs a local search combined with random search for both continuous and combinatorial domains. In this paper, a more robust local searching strategy namely, vantage point procedure is exploited along with random search to deal with complex combinatorial problems. This paper proposes a hybridization technique which involves the Bee Algorithm (BA) and a local search technique based on Vantage Point Tree (VPTs) construction. Following a description of the Vantage Point Bees Algorithm (VPBA), the paper presents the results obtained for several local search strategies for BA, demonstrating efficiency and robustness of the VPBA.

Paper Nr: 57
Title:

L2 Designer - Language and Tool for Generative Art

Authors:

Tomáš Konrády, Barbora Tesařová and Kamila Štekerová

Abstract: We propose a new formal grammar (L2 language) and its implementation in JavaScript tool (L2 Designer). The L2 language allows us to create formal definition of the hierarchy of L-systems encapsulated in L-scripts. The L2 Designer enables creation of initial L-system, its modifications based on genetic programming and iterative evolution, and graphical interpretation. We provide an example of L2 program and we illustrate possibilities of L2 Designer on a case study which was inspired by an artistic decorative floral pattern.

Paper Nr: 59
Title:

A Genetic Algorithm for Training Recognizers of Latent Abnormal Behavior of Dynamic Systems

Authors:

Victor Shcherbinin and Valery Kostenko

Abstract: We consider the problem of automatic construction of algorithms for recognition of abnormal behavior segments in phase trajectories of dynamic systems. The recognition algorithm is trained on a set of trajectories containing normal and abnormal behavior of the system. The exact position of segments corresponding to abnormal behavior in the trajectories of the training set is unknown. To construct recognition algorithm, we use axiomatic approach to abnormal behavior recognition. In this paper we propose a novel two-stage training algorithm which uses ideas of unsupervised learning and evolutonary computation. The results of experimental evaluation of the proposed algorithm and its variations on synthetic data show statistically significant increase in recognition quality for the recognizers constructed by the proposed algorithm compared to the existing training algorithm.

Paper Nr: 60
Title:

Clustering using Cellular Genetic Algorithms

Authors:

Nuno Leite, Fernando Melício and Agostinho Rosa

Abstract: The goal of the clustering process is to find groups of similar patterns in multidimensional data. In this work, the clustering problem is approached using cellular genetic algorithms. The population structure adopted in the cellular genetic algorithm contributes to the population genetic diversity preventing the premature convergence to local optima. The performance of the proposed algorithm is evaluated on 13 test databases. An extension to the basic algorithm was also investigated to handle instances containing non-linearly separable data. The algorithm is compared with nine non-evolutionary classification techniques from the literature, and also compared with three nature inspired methodologies, namely Particle Swarm Optimization, Artificial Bee Colony, and the Firefly Algorithm. The cellular genetic algorithm attains the best result on a test database. A statistical ranking of the compared methods was made, and the proposed algorithm is ranked fifth overall.

Posters
Paper Nr: 6
Title:

Development of an Evolutionary Algorithm for Design of Electron Guns for Material Processing

Authors:

Colin Ribton and Wamadeva Balachandran

Abstract: The design of high quality electron generators is important for a variety of applications including materials processing systems (including welding, cutting and additive manufacture), X-ray tubes for medical, scientific and industrial applications, microscopy, and lithography for integrated circuit manufacture. The many variants of electron gun required, and the increasing demands for highly optimised beam qualities, demands more systematic optimisation methods than offered by trial and error design approaches. This article describes the development of evolutionary algorithms to enable the automatic optimisation of the design of vacuum electron guns. The gun design usually is required to meet specified beam requirements for the applications of interest, so within this work, beam characteristics from the calculated electron trajectories, for example brightness, intensity at focus and beam angle, were derived and used as a measure of the design fitness-for-purpose. Evolutionary parameters were assessed against the efficiency and efficacy of the optimisation process using an analogous design problem. This novel approach offers great potential for producing the next generation of electron guns.

Paper Nr: 11
Title:

Using Genetic Algorithm with Combinational Crossover to Solve Travelling Salesman Problem

Authors:

Ammar Al-Dallal

Abstract: This paper proposes a new solution for Traveling Salesman Problem (TSP) using genetic algorithm. A combinational crossover technique is employed in the search for optimal or near-optimal TSP solutions. It is based upon chromosomes that utilise the concept of heritable building blocks. Moreover, generation of a single offspring, rather than two, per pair of parents, allows the system to generate high performance chromosomes. This solution is compared with the well performing Ordered Crossover (OX). Experimental results demonstrate that, due to the well structured crossover technique, has enhanced performance.

Paper Nr: 14
Title:

Dynamic Feature Selection with Wrapper Model and Ensemble Approach based on Measures of Local Relevances and Group Diversity using Genetic Algorithm

Authors:

Marek Kurzynski, Pawel Trajdos and Maciej Krysmann

Abstract: In the paper the novel feature selection method, using wrapper model and ensemble approach, is presented. In the proposed method features are selected dynamically, i.e. separately for each classified object. First, a set of identical one-feature classifiers using different single feature is created and next the ensemble of features (classifiers) is selected as a solution of optimization problem using genetic algorithm. As an optimality criterion, the sum of measures of features relevance and diversity of ensemble of features is adopted. Both measures are calculated using original concept of randomized reference classifier, which on average acts like classifier with evaluated feature. The performance of the proposed method was compared against six state-of- art feature selection methods using nine benchmark databases. The experimental results clearly show the effectiveness of the dynamic mode and ensemble approach in feature selection procedure.

Paper Nr: 15
Title:

A Reward-driven Model of Darwinian Fitness

Authors:

Jan Teichmann, Eduardo Alonso and Mark Broom

Abstract: In this paper we present a model that, based on the principle of total energy balance (similar to energy conservation in Physics), bridges the gap between Darwinian fitness theories and reward-driven theories of behaviour. Results show that it is possible to accommodate the reward maximization principle underlying modern approaches in behavioural reinforcement learning and traditional fitness approaches. Our framework, presented within a prey-predator model, may have important consequences in the study of behaviour.

Paper Nr: 17
Title:

Weighting and Sampling Data for Individual Classifiers and Bagging with Genetic Algorithms

Authors:

Sašo Karakatič, Marjan Heričko and Vili Podgorelec

Abstract: An imbalanced or inappropriate dataset can have a negative influence in classification model training. In this paper we present an evolutionary method that effectively weights or samples the tuples from the training dataset and tries to minimize the negative effects from innaprotirate datasets. The genetic algorithm with genotype of real numbers is used to evolve the weights or occurrence number for each learning tuple in the dataset. This technique is used with individual classifiers and in combination with the ensemble technique of bagging, where multiple classification models work together in a classification process. We present two variations – weighting the tuples and sampling the classification tuples. Both variations are experimentally tested in combination with individual classifiers (C4.5 and Naive Bayes methods) and in combination with bagging ensemble. Results show that both variations are promising techniques, as they produced better classification models than methods without weighting or sampling, which is also supported with statistical analysis.

Paper Nr: 18
Title:

Combining Development and Evolution - Case Study: One Dimensional Bin-packing

Authors:

Christopher Rajah and Nelishia Pillay

Abstract: The literature highlights the effectiveness of emulating processes from nature to solve complex optimization problems. Two processes in particular that have been investigated are evolution and development. Evolution is achieved by genetic algorithms and the developmental approach was introduced to achieve development. The developmental approach differs from other metaheuristics in that it does not explore the search space applying intensification and diversification to a complete candidate solution. Instead intensification and diversification are performed incrementally, at each step in the process of creating a solution. This is based on an analogy from nature in which a multicellular organism is created incrementally rather than firstly being completely developed and then improved to be fitter. Evolution on the other hand is used to explore the space by applying intensification and diversification to randomly created candidate solutions with the aim of improving the fitness of these candidate solutions and ultimately producing a solution to the problem. Given that in nature once an organism is initially developed its development or growth does not stop at that point but certain cells may continue to grow until a certain point in an organism’s life span, it was felt that the developmental approach terminated prematurely. It was hypothesized that a combination of both these processes, instantiated with development and followed by evolution, would better emulate the processes in nature and would be more effective at exploring the search space. The objective of the research presented in the paper is to test this hypothesis. In terms of search this would mean combining a metaheuristic that applies intensification and diversification incrementally at each step on partial solutions to create initial candidate solutions which are then further explored by a metaheuristic that explores the space of complete candidate solutions. The one-dimensional bin-packing problem was used as a case study to evaluate these ideas. The hybridization of the developmental approach and genetic algorithm was found to perform better than each of these approaches applied separately to solve the problem instances. This study was an initial attempt to test the above hypothesis and has highlighted the potential of this hybridization. Given this future work will apply this approach to other combinatorial optimization problems.

Paper Nr: 22
Title:

Cartesian Genetic Programming in a Changing Environment

Authors:

Karel Slany

Abstract: Evolutionary algorithm are prevalently being used in static environments. In a dynamically changing environment an evolutionary algorithm must be also able to cope with the changes of the environment. This paper describes an algorithm based on Cartesian Genetic Programming (CGP) that is used to design and optimise a solution in a simulated symbolic regression problem in a changing environment. A modified version of the Age-Layered Population Structure (ALPS) algorithm is being used in cooperation with CGP. It is shown that the usage of ALPS can improve the performance on of CGP when solving problems in a changing environment.

Paper Nr: 23
Title:

Elliptical and Archimedean Copulas in Estimation of Distribution Algorithm with Model Migration

Authors:

Martin Hyrš and Josef Schwarz

Abstract: Estimation of distribution algorithms (EDAs) are stochastic optimization techniques that are based on building and sampling a probability model. Copula theory provides methods that simplify the estimation of a probability model. An island-based version of copula-based EDA with probabilistic model migration (mCEDA) was tested on a set of well-known standard optimization benchmarks in the continuous domain. We investigated two families of copulas – Archimedean and elliptical. Experimental results confirm that this concept of model migration (mCEDA) yields better convergence as compared with the sequential version (sCEDA) and other recently published copula-based EDAs.

Paper Nr: 24
Title:

Evolving Four Part Harmony using a Multiple Worlds Model

Authors:

Marco Scirea and Joseph Alexander Brown

Abstract: This application of the MultipleWorlds Model examines a collaborative fitness model for generating four part harmonies. In this model we have multiple populations and the fitness of the individuals is based on the ability of a member from each population to work with the members of other populations. We present the result of two experiments: the generation of compositions, given a static voice line, both in a constrained and unconstrained harmonic framework. The remaining three voices are evolved using this collaborative fitness function, which looks for a number of classical composition rules for such music. The evolved music is found to meet with the requirements placed on it by musical theory. Using the data obtained while running our experiments we observe and discuss interesting qualities of the solution space.

Paper Nr: 33
Title:

Metaheuristic Coevolution Workflow Scheduling in Cloud Environment

Authors:

Denis Nasonov, Mikhail Melnik, Natalya Shindyapina and Nikolay Butakov

Abstract: Today technological progress makes scientific community to challenge more and more complex issues related to computational organization in distributed heterogeneous environments, which usually include cloud computing systems, grids, clusters, PCs and even mobile phones. In such environments, traditionally, one of the most frequently used mechanisms of computational organization is the Workflow approach. Taking into account new technological advantages, such as resources virtualization, we propose new coevolution approaches for workflow scheduling problem. The approach is based on metaheuristic coevolution that evolves several diverse populations that influence each other with final positive effect. Besides traditional population, that optimizes tasks execution order and task's map to the computational resources, additional populations are used to change computational environment to gain more efficient optimization. As a result, proposed scheduling algorithm optimizes both computation tasks to computation environment and computation environment to computation tasks, making final execution process more efficient than traditional approaches can provide.

Paper Nr: 39
Title:

Evolutionary Nonlinear Model Output Statistics for Wind Speed Prediction using Genetic Programming

Authors:

Kisung Seo and Byeongyong Hyeon

Abstract: Wind speed fluctuates heavily and affects a smaller locality than other weather elements. Wind speed is heavily fluctuated and quite local than other weather elements. It is difficult to improve the accuracy of prediction only in a numerical prediction model. An MOS (Model Output Statistics) technique is used to correct the systematic errors of the model using a statistical data analysis. Most previous MOS (Model Output Statistics) used a linear regression model, but they are hard to solve nonlinear natures of the weather prediction. In order to solve the problem of a linear MOS, a nonlinear compensation technique based on evolutionary computation is introduced as a new attempt. We suggest a nonlinear regression method using GP (Genetic Programming) based symbolic regression to generate an open-ended nonlinear MOS. The new nonlinear MOS can express not only nonlinearity much more extensively by involving all mathematical functions, including transcendental functions, but also unlimited orders with a dynamic selection of predictors due to the flexible tree structure of GP. We evaluate the accuracy of the estimation by GP based nonlinear MOS for the three days wind speed prediction for Korean regions. The training period of 2007- 2009, 2011 year is used, the data of 2012 year is for verification, and 2013 year is adopted for test. This method is then compared to the linear MOS and shows superior results.

Paper Nr: 43
Title:

A Flexible and Simplified 2D Environment for Evolving Autonomous Virtual Creatures

Authors:

Ricardo Sisnett

Abstract: In this paper we present a method for creating two-dimensional virtual creatures. Their shape and controlling systems are generated automatically by the use of a genetic algorithms. Unlike previous work, our system has an emphasis in approachability and simplicity, but sacrifices simulation realism. This trade off is done with the intention of using the framework for highly interactive applications such as video games or exhibits.

Paper Nr: 50
Title:

Stabilization of Inverted Pendulum System using Intelligent Linear Quadratic Regulator Controller

Authors:

Salawudeen A. Tijani and M. B. Mua’zu

Abstract: One of the classical problem in dynamics and control theory, which has being widely used as a benchmark for testing control algorithms, such as Linear Quadratic Regulator (LQR) is the balancing of inverted pendulum. The performance of LQR depends largely on the design choice of state and control weighting matrices (Q & R). However, these matrices are usually selected by the designer through a trial and error iterative process which might not guarantee robustness and may increase computational time. To overcome this, we propose a new approach for the optimal determination of the LQR weighting matrices based on weighted artificial fish swarm algorithm (wAFSA). The designed controller is then used to obtain an optimal controller for a dynamic nonlinear Quadruple Inverted Pendulum (QIP). In this paper, we first introduce an approach called inertial weight into the standard Artificial Fish Swarm Algorithm (AFSA) to adaptively select its parameters (visual & step sizes) thereafter, the modified algorithm was used to determine the optimize values of LQR weighting matrices randomly. The optimized values of the weighting matrices were also determined using the standard AFSA and the standard Artificial Bee Colony (ABC) algorithm. This was then used to stabilize the QIP system. Simulation results showed that the proposed method is efficient in determining the weighting matrices of LQR and minimizes time-to-solution in comparison with the conventional trial-&-error approach.

Paper Nr: 52
Title:

Design of a Real Coded GA Processor

Authors:

A. Tsukahara and A. Kanasugi

Abstract: Real Coded Genetic Algorithm (RCGA) has been attracting attention as one of the GA for handling real-valued vectors. Various GA hardware have been proposed, for evolvable hardware, and for an increase in computational throughput. Yet, there are few reports of RCGA hardware. Herein, we propose a design for a real coded GA processor. The proposed processor is implemented using the JGG (Just Generation Gap) as a generation alternation model and the REX (Real coded Ensemble Crossover) as a crossover. In addition, the evaluation functions that depend on problem are calculated using soft macro CPU. The proposed processor is to be applied expected in embedded field applications because of it can be implemented in one chip FPGA.

Paper Nr: 55
Title:

Discovering Internal Fraud Models in a Stream of Banking Transactions

Authors:

Fabien Vilar, Marc Le Goc, Philippe Bouche and Pierre-Yves Rolland

Abstract: Internal frauds in the banking industry represent a huge cost and this problem is particularly difficult to solve because, by construction, swindlers being very imaginative persons, the fraud schemata evolves continuously. Fraud detection systems must then learn from the continuously new fraud schematas, making them difficult to design. This paper proposes a new theoretical and practical approach to detect internal frauds and to model fraud schematas. This approach is based on a particular method of abstraction that reduces the complexity of the problem from O(n2) to O(n) making its implementation in a an Java program that detects and models the frauds in real time and online with a simple professional personal computer. The results of this program are presented with its application on a real-world fraud provided by a world wide French bank.