Séminaires prochains

Séminaires passés

ASEME: Model-driven Software Engineering for Multi-Agent Systems

Nikolaos Spanoudakis (Technical University of Crete, Greece) 0A07 (Pentagone, Plaine de Nimy), le 5/2/2018 à 12h15

Résumé:
The research field of autonomous agents and multi-agent systems was established more than a decade ago as discipline combining ideas from computer science and artificial intelligence. Multi-agent systems are computerized systems composed of multiple intelligent agents interacting within an environment. Multi-agent systems can be used to solve problems that are difficult or impossible to solve by an individual agent or a monolithic system alone. In order to face the complex challenge of developing multi-agent software systems, they need to be integrated into an agent-oriented software engineering (AOSE) methodology. We present ASEME, a hybrid model-driven approach for AOSE. We explain how it improves upon earlier approaches such as MaSE, Gaia, and Tropos.  ASEME's added value is the uniform representation of intra-agent and inter-agent control using a modelling language (statecharts) allowing for seamless integration of an agent's capabilities.

In-memory Stream Indexing and Search of Massive and Fast Incoming Multimedia Content

Dimitris Rafailidis (Aristotle University of Thessaloniki, Grèce) Vésale 025, Plaine de Nimy, le 26/1/2017 à 12h15

Résumé:
The imperative need to process massive multimedia content in a distributed manner has emerged from the proliferation of multimedia applications. Modern multimedia applications in social media platforms, such as Flickr, Instagram and so on, not only have to cope with the vast amount of multimedia content but also with the rate that they are generated. In recognition that billions of images are stored into distributed databases in different logical or physical locations, I will present a new similarity search strategy in a cloud infrastructure based on the notion of "dimensions' value cardinalities" of image descriptors. This new strategy has low preprocessing requirements, by dividing the computational cost of the preprocessing steps into several nodes over the cloud and supports dynamical update/indexing in real-time. Also, I will present an efficient query processing algorithm, where the computation effort is divided into several nodes over the cloud infrastructure. In the second part of my talk I will present some practical issues when building a new infrastructure for multimedia applications. Multimedia search services need to provide an efficient streaming index/storage mechanism, while maintaining high search accuracy. In practice though, some key factors have to be considered such as scalability, latency and robust communication when implementing a stream indexing mechanism in a distributed manner. In this talk, I will introduce a media storm indexing mechanism, where media storms are defined as fast incoming batches. I will show how to index/store massive image collections with varying incoming image rate, using Flink, a distributed in-memory stream processing framework for Big Data.

Live programming the behavioral layer of robots

Par : Miguel Campusano (University of Chile) Salle des Conseils, Plaine de Nimy, le 20/10/2016 à 12h15

Résumé:
The development of robots is a discipline normally detached from the field of computer science. Because of this, the advances in computing sciences and software engineering are not being passed down to the development of robotics software. This can be witnessed by the technical complexity of developing robotics programs, from the installation of the necessary libraries to the programming itself and even the deployment of the programs on the robots. At the University of Chile we are working to integrate software engineering techniques and studies in the development of robotics software. In this talk I will present you an overview of our work about software engineering in the world of robotics. In particularly, I will present you the development of a new language for the behavioral layer of robots: LRP. LRP uses live programming, a technique that allows programmers to benefit from an immediate connection with the program that they are making thanks to an immediate, 'live' feedback. In LRP, this allows for extremely rapid creating, or variation, of robot behavior and for dramatically increased debugging speed.

Quand l'informatique réinvente la sociologie !

Par : Hugues Bersini (ULB) 0A11/P, le 21/1/2016 à 16h00 Hugues Bersini est professeur d’informatique et directeur du laboratoire d’intelligence artificielle à l’Université libre de Bruxelles. Ses travaux de recherche couvrent l’intelligence artificielle, la bioinformatique, le génie logiciel, les systèmes complexes et les sciences cognitives. Il est Membre de l’Académie royale de Belgique.

Résumé:
Le tout social est plus lent, plus fragile, plus raciste, plus communautariste, plus délétère et plus inégal que les individus qui le constituent. Si les pères de la sociologie, Durkheim, Comte, avaient disposé d’un ordinateur à leur époque, voilà sans doute le type de sociologie informatisée qu’ils auraient préféré inventer pour analyser ces réalités.
Voir aussi son exposé, le même jour dans le cadre des Jeudis du libre.

Preferences, Queries, Logics

Par : Jan Chomicki (University of Buffalo) 0A11/P, le 8/12/2015 à 12h15 Jan Chomicki (http://www.cse.buffalo.edu/~chomicki/) received his M.S. in computer science with honors from Warsaw University, Poland, and his Ph.D., also in computer science, from Rutgers University. He is currently a Professor in the Department of Computer Science and Engineering, University at Buffalo, State University of New York. He has held visiting positions at Hewlett-Packard Labs, European Community Research Center, Bellcore, Lucent Bell Labs, Warsaw University, and Free University of Bozen-Bolzano. He is the recipient of 7 National Science Foundation research grants, a Fulbright Lecturer, an Erasmus Mundus Scholar, and the author of over 90 research publications. He has served on the editorial boards of ACM Transactions on Database Systems, Information Processing Letters, Journal of Applied Logic, and Knowledge and Information Systems. He has given many invited talks in academia and industry, and keynote addresses and tutorials at international database conferences. He was the General Co-Chair of the 2003 TIME-ICTL Symposium and the Program Co-Chair of TIME 2005. He has also served on the program committees of SIGMOD, PODS, VLDB, ICDT, ICDE, EDBT, and many other conferences and workshops. He has co-edited two books: "Logics for Databases and Information Systems" (Kluwer, 1998) and "Logics for Emerging Applications of Databases" (Springer, 2003). His research addresses issues in logical foundations of databases, data integrity, and query languages.

Résumé:
The notion of preference plays an increasing role in today's information systems. In particular, preferences are used to specify which query answers are the best from the user's point of view. In this talk, we survey selected work done over the last 10 years in the context of preference queries to relational database systems. We focus on preferences that are specified logically. We show how such preferences can be embedded into relational query languages. By separating the preferences from the query, preference-specific query evaluation and optimization techniques can be formulated. We also study the foundations of skyline queries - the most popular class of preference queries. Finally, we outline a personal view on the future prospects of preference research.

Téléopération du robot NAO avec le capteur "Kinect for Windows V2"

Par : Anthony Rouneau (Etudiant en sciences informatiques, UMONS) Salle des conseils, le 29/10/2015 à 12h15 Résumé: Dans cet exposé, j'exposerai le résultat du travail réalisé cet été dans le cadre d'un stage de recherches effectué dans le service de Génie Logiciel de l'UMONS. Ils s'agit pour un robot humanoïde de type NAO, équipé d'une Kinect V2, de reproduire les mouvements effectués par un être humain se trouvant en face de lui. Après avoir expliqué la solution mise en place pour résoudre le problème, une démonstration sera réalisée. Egalement, je parlerai des capacités et des limitations du robot ainsi que celles de la Kinect. Je tâcherai également de répondre aux questions suivantes : peut-on compter sur l'équilibre du robot ? Quelles données de la Kinect utiliser pour ce genre d'opération ?

Distances d'édition entre permutations et graphes

Par : Anthony Labarre Salle des conseils, le 1/10/2015 à 12h15 Anthony Labarre est professeur associé en Informatique à l'Université de Paris-Est Marne-la-Vallée. Il est membre du groupe de recherche en Algorithmique du LIGM (Laboratoire d'Informatique Gaspard-Monge). Ses recherches portent sur le réarrangement de génomes, sur la compression de données et le calcul d'ADN.

Résumé:
Les distances d'édition quantifient les différences entre les objets qu'elles comparent en termes de nombre d'opérations élémentaires à appliquer pour transformer un objet en un autre, étant entendu que les transformations autorisées sont fixées au préalable. La nature des objets à comparer et le choix des opérations utilisables donnent lieu à une multitude de questions algorithmiques et mathématiques, et le cas particulier où les objets à comparer sont des permutations (éventuellement signées) trouve des applications dans de nombreux domaines comme la bioinformatique, la conception de réseaux d'interconnexion, et l'agrégation de classements. Je présenterai un tour d'horizon de plusieurs problèmes de cette famille, ainsi que des techniques mises en oeuvre pour les résoudre exactement ou approximativement, issues pour la plupart de la théorie des graphes.