Description
For the first time, an EGEE User Forum will feature a general Computer Science session. Indeed, many contributions come from Computer scientists that exploit the grid to run large-scale computations, for Machine Learning, Optimization, Game theory, and Complex networks. Phenomenology of the algorithms, theoretical results based on large experiments, and inter-disciplinary applications will be described, at various levels form mature tools to research-focused work. The Grid Observatory Cluster of EGEE-III will also present recent advances in exploration and analysis tools that are oriented towards the scientific view of grids. The session will be an opportunity to:
Dr
Philippe Gauron
(LRI)
4/14/10, 11:00 AM
End-user environments, scientific gateways and portal technologies
Oral
Reaping the full benefit of the Grid Observatory (GO) initiative requires providing end-users with convenient representation of the traces. The present lack of standardization creates considerable difficulties for developing automated analysis and situation handling solutions. We report an experiment on the representation of internal logs of the gLite WMS with the IBM Common Base Event (CBE)...
Mr
Lovro Ilijasic
(University of Turin)
4/14/10, 11:20 AM
Scientific results obtained using distributed computing technologies
Oral
Large-scale analysis of grid job data, consisting of more than 28 million jobs gathered during 20 months from all major EGEE Resource Brokers, provides interesting results regarding the properties of job time parameters. The main parts in a job’s life cycle are observed to have different distributions, having also different origins that cause these behaviors. While ‘match time’ parameter...
Ms
Fani Tzima
(Aristotle University of Thessaloniki), Mr
Fotis Psomopoulos
(Aristotle University of Thessaloniki)
4/14/10, 11:40 AM
Scientific results obtained using distributed computing technologies
Oral
Strength-based Learning Classifier Systems (LCS) are machine learning systems designed to tackle both sequential and single-step decision tasks by coupling a gradually evolving population of rules with a reinforcement component. ZCS-DM, a Zeroth-level Classifier System for Data Mining, is a novel algorithm in this field, recently shown to be very effective in several benchmark classification...
Mr
Fotis Psomopoulos
(Aristotle University of Thessaloniki), Mr
Kyriakos Chatzidimitriou
(Aristotle University of Thessaloniki)
4/14/10, 12:00 PM
Scientific results obtained using distributed computing technologies
Oral
In this work we use the NeuroEvolution of Augmented Topologies (NEAT) methodology, for optimising Echo State Networks (ESNs), in order to achieve high performance in machine learning tasks. The large parameter space of NEAT, the many variations of ESNs and the stochastic nature of evolutionary computation, requiring many evaluations for statistically valid conclusions, promotes the Grid as a...
Dr
Ruediger Berlich
(Steinbuch Centre for Computing, Karlsruhe Institute of Technology)
4/14/10, 2:00 PM
Experiences from application porting and deployment
Oral
The Geneva library implements parallel and distributed parametric optimization algorithms, capable of running on devices ranging from multi-core systems over clusters all the way to Grids and Clouds. The generic design makes Grid-resources available to user groups that have formerly not been exposed to such environments.
Mr
Adam Padee
(Warsaw University of Technology)
4/14/10, 2:20 PM
Scientific results obtained using distributed computing technologies
Oral
This paper describes an optimization framework which utilizes a distributed evolutionary algorithm with the ability of adaptation to the complex physical structure of the grid fabric. The algorithm is divided into several partitions (demes) which are located on physical clusters. This architecture allows simultaneous use of all the available resources, regardless of their geographical...
Mr
Karol Wawrzyniak
(Interdisciplinary Centre for Mathematical and Computational Modelling, University of Warsaw)
4/14/10, 2:40 PM
Scientific results obtained using distributed computing technologies
Oral
We present a comprehensive study of the utility function of the minority game in its efficient regime. We develop an effective description of the state of the game. For both the step-like payoff function g(x) = sgn(x) and the proportional function g(x)=x, we explicitly represent the game as the Markov process and prove the finiteness of number of states. We also demonstrate boundedness of the...
Prof.
Panos Argyrakis
(Physics Department, Aristotle University of Thessaloniki), Mr
Paschalis Korosoglou
(Grid Operations and HPC Centre, Aristotle University of Thessaloniki)
4/14/10, 3:00 PM
Scientific results obtained using distributed computing technologies
Oral
We use Monte-Carlo simulations to study the dynamics of the irreversible A+B->2A two species reaction on regular lattices, Erdos-Renyi (ER) and scale-free (SF) networks. The problem we study is an analogue to the spread of a virus in computer networks and other epidemiological models, such as information diffusion (i.e. rumor spreading) in social networks, and information propagation among...