Free attendance, mandatory registration - Please register before September 16th
Alfred O. Hero III is a Professor in three departments of the University of Michigan, Ann Arbor, namely the department of Electrical Engineering and Computer Science, the department of Biomedical Engineering and the Department of Statistics. His recent research interests include inference in sensor networks, adaptive sensing, bioinformatics, inverse problems, and statistical signal and image processing. He is a Fellow of the Institute of Electrical and Electronics Engineers (IEEE) and has received the IEEE Third Millennium Medal (2000). He was President of the IEEE Signal Processing Society (2006-2008).
A minimal spanning tree (MST) spanning random points has total spanning length that converges to the entropy of the underlying density generating the points. This celebrated result was first established by Beardwood, Halton and Hammersley (1958) and has since been extended to other random Euclidean and non-Euclidean graphs, such as the geodesic MST (GMST) and the k-nearest neighbour graph (kNNG) over a random set of points. Using the BHH theory of random graphs one can construct graph-based estimates of topological properties of a high dimensional distribution of a data sample. This leads, for example, to a model-free consistent estimator of intrinsic dimension of a data manifold and a high performance non-parametric anomaly detector. Alfred Hero’s talk will illustrate this entropic graph approach for applications including anomaly detection in Internet traffic, activity detection in a MICA2 wireless network and intrinsic dimension estimation of image databases.
Peter Clote, Full Professor in Computer Science and Biology, Boston College, studied mathematics at MIT, Duke University and Université Paris VII. He has worked in departments of mathematics, computer science, and biology and led the establishment of a PhD program in bioinformatics that began accepting students in 1999. He has co-written two books: Computational Molecular Biology: An Introduction (2000), and Boolean Functions and Computation Models (2002). He plays alto and soprano sax semi-professionally in the trio Sharp Eleventh and is preparing a CD of jazz originals.
Similar to DNA, ribonucleic acid (RNA) can play a role in transmitting genetic information. This is clearly illustrated in RNA viruses, such as human immunodeficiency virus (HIV) and Hepatitis C virus (HCV), whose genome consists entirely of RNA. However, unlike DNA, single-stranded RNA can form complicated 3-dimensional structures that catalyze essential molecular reactions, such as formation of a peptide bond, necessary in the formation of proteins.
Did life begin with RNA? How does RNA regulate genes, i.e. turn genes “off” and “on”? In this talk, Peter Clote will describe his work on dynamic programming to answer fundamental questions about the folding landscape of RNA molecules, and be used to find riboswitches, portions of messenger RNA having two distinct structures or “conformations”, corresponding to “gene on” and “gene off” states.
Olivier Teytaud is a junior researcher at INRIA centre de Saclay working in the TAO Project on machine-learning and optimization (TAO is a common project between INRIA, Université Paris-Sud and CNRS). His research activity includes reinforcement learning, stochastic dynamic programming and evolutionary algorithms. He is participating to the team developing MoGo, a computer Go team using innovative algorithms based on Monte-Carlo algorithms and advanced planification techniques. He is especially involved in the parallelization of the program. In March 2008, Mogo was the first computer Go program to win against a professional player on a 9x9 board.
Olivier Teytaud’s home page : http://www.lri.fr/~teytaud/cv/cv.html
The project presents the Bandit-Based Monte-Carlo Planning algorithm, and some improvements in terms of introduction of human-expertise, exploration-exploitation innovations, parallelization. The MoGo project applies these works to the very old and very hard game of Go, in which computers are traditionally much weaker than in other games. The project provided the first ever wins against a professional player in 9x9 Go.
Dominique Lesselier is Directeur de recherche at CNRS, with the Laboratoire des Signaux et Systèmes, joint laboratory of Supélec, CNRS, and Université Paris-Sud, Gif-sur-Yvette. He is also Director of the Groupement de Recherche CNRS "Ondes" 2006-2009.
Alain Le Brun is senior scientist at EDF R&D in the Department Simulation and Information Technologies for Power Generation Systems, Chatou.
Thierry Collette belongs to CEA LIST, which is based near Paris. He leads since 2004 the Architectures and Design Unit of LIST, unit of 100 researchers, focused on embedded computing solutions. He has designed several parallel and reconfigurable embedded computers. He wrote, as author and co-author, several papers in conferences and journals and holds several patents. He teaches computer architectures in master degree in different universities.
Joël Monnier is the CEO of the French start-up Kalray. He has been previously Corporate Vice President of ST Microelectronics in charge of R&D from 1989 to 2004. In this function, he has been the main actor for the emergence of the Crolles 2 Microelectronics Center (a joint program between STM, Philips and Motorola). He spent several years of his career (from 1983 to 2004) in various important positions in ST Microelectronics (initially Thomson CSF EFCIS) where he was a key actor to reinforce the competitiveness of ST Microelectronics technology in order to guarantee the strategic independence of the Group. He has promoted and supported ST participation in many cooperative European R&D programs like MEDEA+. Previously, he worked 9 years in various positions in Texas Instrument.
Emergence of new embedded applications and the evolution of microelectronic technologies allow propositions of new computing architectures, such as multicore and reconfigurable architectures. The presentation will deal with these major evolutions, especially focusing on the roadmap of CEA-LIST on embedded computing solutions and on the presentation of the start-up Kalray.
Caroline Appert graduated at University Paris-Sud in 2007 in the in|situ| project (joint project between INRIA and LRI). She is currently doing a post-doc at IBM Almaden Research Center. Her thesis, on modeling, evaluating and generating interaction techniques was awarded by the Gilles Kahn's prize in 2007.
Current industrial applications do not take benefits of Human Computer Interaction innovations. Caroline Appert will present an approach which combines both predictive and empirical aspects to support a generative design for graphical interfaces. Her talk will give an overview of the tools to support this approach and illustrate it by some interaction techniques.
Marc Pouzet is Professor at University Paris-Sud and a junior member of Institut Universitaire de France. He works in the Laboratoire de Recherche en Informatique in a joint Project ProVal with the INRIA Centre de Recherche Saclay - Île-de-France. His research activity concerns languages for real-time systems with particular interest in synchronous programming, semantics, type-systems and compilation.
Serge Abiteboul is a senior researcher at INRIA Centre de Recherche Saclay - Île-de-France working in the Gemo project on Integration of Data and Knowledge Distributed over the Web, a joint project with Université Paris-Sud and CNRS. Serge Abiteboul is a world-wide recognised expert in database systems. He received the 1998 ACM SIGMOD Innovation Award and obtained the EADS award from Académie des Sciences in 2007. He is the program chair of VLDB09 (the International Conference on Very Large Databases, the premier forum for research on data management).
Serge Abiteboul’s home page : http://www-rocq.inria.fr/~abitebou/
Information ubiquity has created a large crowd of users (e.g. scientists), who can use DBMS technology to process and share their data more effectively. We consider the issue of building content sharing communities. The users should be able to manage and share their data with minimal effort with the system in charge of indexing it (to make it accessible), replicating it (for availability), and reorganizing its physical storage (for better query performance). Serge Abiteboul will outline research challenges that must be addressed in order to realize this vision.
Jean-Luc Gauvain is a senior scientist at the CNRS and head of the Spoken Language Processing Group at LIMSI-CNRS. His primary research interests are on large vocabulary continuous speech and audio indexing including conversational interfaces, speaker recognition, language identification, and speech translation. He was awarded a CNRS silver medal in 2007. He was a member of the IEEE Signal Processing Society's Speech Technical Committee from 1998 to 2001. He is currently co editor-in-chief of the Speech Communication journal.
Jean-Luc Gauvain’s home page : http://www.limsi.fr/~gauvain/
With the expansion of information sources, there is a pressing need for automatic processing of audio data in multiple languages. Many applications could greatly benefit from spoken language technologies such as the creation and access to digital multimedia libraries, audio-visual content-based indexation, media monitoring, News-on-Demand and Internet watch services. Automatic speech recognition and translation of spoken data are key technologies for such applications.
José Lozada was awarded the degree of Master of Science in Robotics from ENS in Paris in 2003 and his Ph.D. from the Ecole Polytechnique in Paris in 2007. He joined the Sensory Interface laboratory at CEA-LIST as a project manager in the field of mechatronics of the human-machine interfaces. He is the co-inventor of 4 patents and the co-author of 10 scientific publications in the field of haptics and smart materials. His research interests lie in the field of robotics, haptic and tactile interfaces, and smart materials.
The main objective of this research work is to develop a novel digital keyboard capable of reproducing the behaviour of a grand piano action mechanism. The design and the control of a novel haptic interface based on magneto-rheological fluids are presented. The modelling and the identification of an original operation mode for MR fluids is presented and used for the design of the haptic interface for musical keyboards. An analytical model of the interface key is built and used to develop the control law. The mechatronic model is numerically simulated and compared to the real behaviour of the interface. Finally a real time control loop coupled to a virtual model is used to command the interface (for instance the dynamical model of the traditional piano).
