Best Student Paper Award

Wednesday, 5 May 2010, 09:00-13:00

The conference awarded the best student paper. After a pre-selection round, 7 papers where the primary author is a student have been nominated for the award and where orally presented in front of the public and the best student paper committee. The abstracts of the nominated papers are given here, full texts will be included in the conference proceedings.

Taking into account the presentations as well as the quality of the presented work itself, the committee chose Claudia Lindner's A Market-Affected Sealed-Bid Auction Protocol as the Best Student Paper of SETN 2010. Ms Lindner was presented with the award during the opening ceremony on the same day.

Committee

Chair, Efstathios Stamatatos, University of the Aegean

Manolis Koubarakis, National and Kapodistrian University of Athens
Aristidis Likas, University of Ioannina
Giorgos Stamou, National Technical University of Athens
Maria Virvou, University of Piraeus

Anabela Moreira Bernardino, Eugénia Moreira Bernardino, Juan Manuel Sánchez Pérez, Juan Antonio Gómez Pulido and Miguel Angel Vega Rodríguez,
A Hybrid Ant Colony Optimization Algorithm for Solving the Ring Arc-Loading Problem

Abstract: The past two decades have witnessed tremendous research activities in optimization methods for communication networks. One important problem in communication networks is the Weighted Ring Arc-Loading Problem (combinatorial optimization NP-complete problem). This problem arises in engineering and planning of the Resilient Packet Ring (RPR) systems. Specifically, for a given set of non-split and uni-directional point-to-point demands (weights), the objective is to find the routing for each demand (i.e., assignment of the demand to either clockwise or counter-clockwise ring) so that the maximum arc load is minimised. In this paper, we propose a Hybrid Ant Colony Optimization Algorithm to solve this problem. We compare our results with the results obtained by the standard Genetic Algorithm and
Particle Swarm Optimization, used in literature.

Giorgos Giannakouris, Vassilios Vassiliadis, and George Dounias,
Experimental Study on a Hybrid Nature-Inspired Algorithm for financial portfolio optimization

Abstract: Hybrid intelligent schemes have proven their efficiency in solving NP-hard optimization problems. Portfolio optimization refers to the problem of finding the optimal combination of assets and their corresponding weights which satisfies a specific investment goal and various constraints. In this study, a hybrid intelligent metaheuristic, which combines the Ant Colony Optimization algorithm and the Firefly algorithm, is proposed in tackling a complex formulation of the portfolio management problem. The objective function under consideration is the maximization of a financial ratio which combines factors of risk and return. At the same time, a hard constraint, which refers to the tracking ability of the constructed portfolio towards a benchmark stock index, is imposed. The aim of this computational study is twofold. Firstly, the efficiency of the hybrid scheme is highlighted. Secondly, comparison results between alternative mechanisms, which are incorporated in the main function of the hybrid scheme, are presented.

Vasilis Giannopoulos and Pavlos Peppas,
Associations between Constructive Models for Set Contraction

Abstract: Belief Change is one of the central research topics in Knowledge Representation and theory revision and contraction are two of the most important operators in Belief Change. Recently the original axiomatization of revision and contraction was extended to include epistemic input represented by a (possibly infinite) set of sentences (as opposed to a single sentence) giving rise to the operators of set revision (also known as multiple revision) and set contraction. Both set revision and set contraction have been characterized in terms of constructive models called system of spheres and epistemic grasp respectively. Based on these links, in this paper we provide a characterization of set contraction in terms of system of spheres, and we identify the necessary and sufficient conditions under which the system-of-spheres model and the epistemic-grasp model give rise to the same set contraction.

Claudia Lindner,
A Market-Affected Sealed-Bid Auction Protocol

Abstract: Multiagent resource allocation defines the issue of having to distribute a set of resources among a set of agents, aiming at a fair and efficient allocation. Resource allocation procedures can be evaluated with regard to properties such as budget balance and strategy-proofness. Designing a budget-balanced and strategy-proof allocation procedure that always provides a fair (namely, envy-free) and efficient (namely, Pareto-optimal) allocation poses a true challenge. To the best of our knowledge, none of the existing procedures combines all four properties. Moreover, in previous literature no attention is given to the allocation of unwanted resources (i.e., resources that seem to be of no use for all agents) in a way as to maximize social welfare. Yet, dealing inappropriately with unwanted resources may decrease each agent's benefit. Therefore, we extend the scope of sealed-bid auctions by means of involving market prices so as to always provide an optimal solution under consideration of each agent's preferences. We present a new market-affected sealed-bid auction protocol (MSAP) where agents submit sealed bids on indivisible resources, and we allow monetary side-payments. We show this protocol to be budget-balanced and weakly strategy-proof, and to always provide an allocation that maximizes both utilitarian and egalitarian social welfare, and is envy-free and
Pareto-optimal.

Vangelis Oikonomou and Konstantinos Blekas,
A sparse spatial linear regression model for fMRI data analysis

Abstract: In this study we present an advanced Bayesian framework for the analysis of functional Magnetic Resonance Imaging (fMRI) data that simultaneously employs both spatial and sparse properties. The basic building block of our method is the general linear model (GML) that constitute a well-known probabilistic approach for regression. By treating regression coefficients as random variables, we can apply an appropriate Gibbs distribution function in order to capture spatial constraints of fMRI time series. In the same time, sparse properties are also embedded through a RVM-based sparse prior over coefficients. The proposed scheme is described as a maximum a posteriori (MAP) approach, where the known Expectation Maximization (EM) algorithm is applied offering closed form update equations. We have demonstrated that our method produces improved performance and enhanced functional activation detection in both simulated data and real applications.

Theodore Patkos, Ioannis Chrysakis, Antonis Bikakis, Dimitris Plexousakis, and Grigoris Antoniou,
A Reasoning Framework for Ambient Intelligence

Abstract: Ambient Intelligence is an emerging discipline that requires the integration of expertise from a multitude of scientific fields. The role of Artificial Intelligence is crucial not only for bringing intelligence to everyday environments, but also for providing the means for the different disciplines to collaborate. In this paper we describe the design of a reasoning framework, applied to an operational Ambient Intelligence infrastructure, that combines rule-based reasoning with reasoning about actions and causality on top of ontology-based context models. The emphasis is on identifying the limitations of the rule-based approach and the way action theories can be employed to fill the gaps.

Sotiris Tasoulis, Charalampos Doukas, Ilias Maglogiannis, and Vassilis Plagianakos,
Skin Lesions Characterisation Utilising Clustering Algorithms

Abstract: In this paper we propose a clustering technique for the recognition of pigmented skin lesions in dermatological images. It is known that computer vision-based diagnosis systems have been used aiming mostly at the early detection of skin cancer and more specifically the recognition of malignant melanoma tumour. The feature extraction is performed utilising digital image processing methods, i.e. segmentation, border detection, colour and texture processing. The proposed method belongs to a class of clustering algorithms which are very successful in dealing with high dimensional data, utilising information driven by the Principal Component Analysis. Experimental
results show the high performance of the algorithm against other methods of the same class.