Palestrante: Atze Dijkstra
Department of Information and Computingt Sciences da Utrecht University - The Netherlands Data: 10/10/2008 Horário: 15:00 Local: Sala 2077 Resumo: The Essential Haskell Compiler (EHC) is a Haskell compiler for use in research and education. As a language it features
language extensions such as higher-ranked types. As a system it features a decomposition into smaller subsystems that facilitates experimentation. In this talk we look at some of the language extensions and how the system is organised to cope with the inherent complexity of a compiler for a large language like Haskell. Our approach splits both language and implementation into smaller, manageable steps, and uses specific tools to generate parts of the compiler from higher level descriptions. In particular we use attribute grammars and domain specific languages for type systems as well as variant (de)composition.
Referência sobre EHC: http://www.cs.uu.nl/wiki/Ehc/WebHome
Referência sobre o palestrante: http://www.cs.uu.nl/~atze
Seminário: Mathematics and computer science serving/impacting Bioinformatics
Palestrante: Prof. Gaston Gonnet
ETH Zurich Switzerland Data: 21/08/2008 Horário: 14:00 Local: Sala 2077 Resumo: Since the early days of Bioinformatics, it was clear that mathematics in general, and computer science in particular have a lot to contribute to bioinformatics. Bioinformatics has made substantial progress in the last 20 years, using tools from computer science, mathematics and statistics. This progress has been rather explosive and in many cases suboptimal when observed from a traditional mathematical eye. We will present the areas where we consider that there is still plenty of room for contributions which touch computational mathematics. In particular the areas of modelling and simulation, maximum likelihood methods and the interface between bioinformatics and computer algebra.
Seminário: Modelling Mobile Robots Through System Identification
Palestrante: Prof. Ulrich Nehmzow
Ulrich Nehmzow is Professor of Cognitive Robotics at the University of
Ulster, and leads the cognitive robotics group there. He obtained a
Diplom in Electrical Engineering and Data: 28/03/2008 Horário: 14:00 Local: 2077 Resumo: The operation of robots in "real world" scenarios, i.e. under the
influence of noise, variation and unpredictablity, is highly complex,
usually non-linear and therefore hard to model. Modelling approaches
such as artificial neural network modelling can represent input-output
relationships accurately, but have the disadvantage that the resulting
models are opaque and therefore not amenable to mathematical analysis.
They are also often computationally expensive, which can make them
unsuitable for performance-critical or safety-critical applications
(e.g. space exploration, operating in hazardous environments etc).
However, transparent, analysable models can be obtained through Narmax
system identification. These models are given in closed mathematical
form, and they are parsimonious, which makes them suitable for direct
robot control.
In this talk I will describe the approach taken in the recent RobotMODIC
project, and show experimental results in automated robot programming,
robot modelling and environment modelling, using Magellan Pro and Scitos
G5 mobile robots.
Further information can be found at
http://infm.ulst.ac.uk/~ulrich/RobotMODIC/
Seminário: Coordenação minimalista de um time de robôs usando redes: um estudo de caso
Palestrante: Marcos Vieira, University of Southern California, EUA
Data: 06/03/2008 Horário: 14:00 Local: Sala 2077 - ICEx Resumo: Nós desenvolvemos e analisamos experimentalmente um sistema com vários robôs móveis. Um time de robôs (chamaremos de polícia) tenta capturar um outro time de robôs (chamaremos de ladrão). O objetivo do jogo é capturar todos os robôs ladrões. O time de policias tem a sua disposição uma rede de sensores embutida no ambiente. O problema é uma instância de coordenação entre vários robôs, onde a rede de sensores aumenta o poder de comunicação e observação dos robôs. Nós discutimos as decisões de projeto no desenvolvimento do sistema com ênfase na natureza minimalista dos robôs. Nós desenvolvemos um simulador para guiar na escolha dos experimentos (foi escolhido o pior caso). Os experimentos foram feitos no mundo real, onde tanto os robôs como a rede de sensores são estressados. Surpreendentemente, os resultados mostram que os experimentos são próximos da simulação mesmo com nossas decisões de projeto minimalista.
Links:
Projeto
http://enl.usc.edu/projects/peg/index.html
Plataforma
http://enl.usc.edu/projects/peg/platform.html
Seminário: Pesquisa em redes sem fio e auto-organizáveis do grupo de pesquisa em "internetworking" da UCSC - Universidade da Califórnia Santa Cruz
Palestrante: Profa. Katia Obraczka
Katia Obraczka is currently an Associate Professor of Computer Engineering at the University of California, Santa Cruz. Before joining UCSC, she held a research scientist position Data: 09/08/2007 Horário: 15:00 Local: Sala 2077 Resumo: This talk will highlight some of the research projects on wireless self-organizing networks that are currently underway at the UCSC Internetwork Research Group (i-NRG). These projects include protocol design, evaluation, and implementation at various layers of the protocol stack as well as sensor network system development, implementation, and deployment.
(O seminário será apresentado em português)
Seminário: "Opportunities for CS Graduate Students at the University of Alberta" e "Join Processing in Sensor Networks: Where and How"
Palestrante: Prof. Mario Nascimento
University of Alberta, Canadá Data: 07/08/2007 Horário: 14:00 Local: Sala 2077 Resumo: RECRUITING: Opportunities for CS Graduate Students at the University of Alberta
In this short presentation we will give an overview of the University
of Alberta's Department of Computing Science. The department is on
the most prosperous Canadian province and offers a number of
opportunities for graduate students both at the M.Sc and Ph.D.
- most such opportunities are locally funded through Teaching and/or
Research Assistantships. Opportunities for short term visits,
e.g., "sandwich internships" and/or sabbatical visits will be also
discussed.
TECHNICAL PRESENTATION (DB Oriented): Join Processing in Sensor Networks: Where and How.
Mario A. Nascimento
Univ. of Alberta, Canada
http://www.cs.ualberta.ca/~mn/bio.html
Sensor networks are made of autonomous devices that are able to
collect, store, process and share data with other devices. We
consider the problem of processing join queries in a wireless
sensor network, by decomposing the problem into two complementary
subproblems: where the join can be processed and then how to
process the same. For the first subproblem we propose four
strategies and investigate their performance across several
scenarios. Not surprisingly, our experiments show that no single
strategy performs best for all scenarios. In order to avoid
the potential high cost of using a fixed strategy for processing
all queries, we use a cost-based model to select the best join
strategy for the query at hand. Our experiments confirm that,
given a set of queries, selecting the join strategy based on the
cost model is always better than using any fixed strategy for all
queries. For the second subproblem we propose and investigate
a distributed algorithm for join processing that considers memory
limitations at nodes and does not make any a priori assumptions on
the available memory at the processing nodes. At the same time,
our algorithm still minimizes the energy cost of query processing.
(Joint work w/ Alex Coman and Joerg Sander. Partially funded by NSERC,
Canada)
Seminário: A resilient protocol from 13th century Venice e Why P2P pair-wise reciprocation can scale, after all
Palestrante: Miranda Mowbray
Pesquisadora HPs Research Labs., Bristol, Inglaterra Data: 06/06/2007 Horário: 14:30 Local: Sala 2076 - ICEx Resumo: "A resilient protocol from 13th century Venice"
Abstract:
This talk is about the protocol, designed in 1268, which was used for electing the ruler of Venice. It is very different from modern protocols, and at first sight seems bizarre. However it has some useful properties, that, in addition to being interesting in themselves, may be applicable to the problem of leader election in distributed computer networks. In particular, it offers some opportunities to minorities while ensuring that the most popular candidates are most likely to win; and it offers some resiliance to corruption of voters.
"Why P2P pair-wise reciprocation can scale, after all"
(Joint work with Nazareno Andrade, Francisco Brasileiro, and Walfredo
Cirne).
Abstract:
The Network of Favours is a mechanism used to deter free riding in OurGrid. It works by using pair-wise reciprocation; that is, one peer decides how to treat another based only on its direct experience of interacting with the peer in the past. Previous work by Feldman, Lai and other researchers suggests that no pair-wise reciprocation mechanism for a P2P system can scale well beyond 1,000 peers. However, the Network of Favors actually performs slightly better for 10,000 peers than for 1,000
or 100. I will explain the features of the Network of Favors and OurGrid that make this possible.
Av. Antônio Carlos 6627 - Pampulha
Prédio do ICEx - Anexo U - sala 5309
Belo Horizonte - Minas Gerais - Brasil
CEP: 31270-010
Telefone: (31) 3409-5860 Fax: (31) 3409-5858