Zsofia Lendek (Technical University of Cluj-Napoca, Department of Automation, Roumanie)
« Observer design using Takagi-Sugeno fuzzy models »
Abstract: Since the introduction of Takagi-Sugeno (TS) fuzzy models in 1985, these models have been extensively used for the stability analysis, controller and observer design of nonlinear systems. Depending on the way they are obtained, TS models may be an approximation or exact representation a given nonlinear system. They are a collection of local, usually linear, models blended together by scalar functions, a structure that facilitates the application of the direct Lyapunov method. Stability and design conditions are usually formulated as linear matrix inequalities that can be solved using available convex optimization techniques.
The talk will start with an introduction to TS models and their use for estimation and control. Next, some of the main research problems that have been addressed in observer design in the last decades will be reviewed. Afterwards, I will present some recent results that take TS models even closer to general nonlinear systems. Finally, I will discuss and illustrate the advantages and shortcomings of fuzzy models for estimation.
Gabriella Pasi (University of Milano-Bicocca, Department of Informatics, Systems, and Communication, Italie)
« Assessing Information Credibility in the Social Web »
Résumé: In the context of the Social Web, where a large amount of User Generated Content (UGC) is diffused through Social Media without any form of trusted external control, the risk of running into misinformation is not negligible. For this reason, the issue of assessing the veracity or credibility of “potential” information is of increasing interest and importance, especially in certain domains like health. In the last few years several approaches have been proposed to automatically assess the credibility of UCG in Social Media. Most are data-driven approaches, based on machine learning techniques, but recently model-driven approaches are also being investigated, in particular, approaches relying on the Multi Criteria Decision Making paradigm. This talk will summarize the approaches aimed at tackling this problem are addressed, with particular emphasis on model-driven approaches; their application to specific problems will also be addressed. Moreover, the issue of considering information credibility in accessing information on the Web via Search Engines will be also discussed. In the field of Information Retrieval, the study of how to access relevant information with respect to users’ information needs has been steadily developing over the last fifty years. It has developed from considering only topical relevance of documents, to taking into account the concept of popularity in Web search engines, to considering contextual aspects in personalized search. With the proliferation of misinformation spread through both the Web and social media, a new challenge arises in the IR field: to provide users with information that is both relevant and credible. Depending on whether one searches Web pages or social content, depending on the task and the domain for which the search is performed, the concept of credibility understood as an aspect of relevance may change. Some issues related to credibility assessment in IR, and the problem of defining datasets for experimental evaluation of approaches that address those issues will be discussed.
Grégory Smits (Université de Rennes, IRISA, France)
« Enrichissement des méthodes d’accès aux données »
Résumé: Quasiment toutes nos activités, personnelles ou professionnelles, génèrent des données numériques. Leur analyse peut s’avérer très précieuse pour comprendre nos comportements, optimiser nos déplacements, surveiller notre santé, mais elle est souvent très lucrative pour du démarchage commercial ciblé. La transformation de données, brutes ou structurées, en connaissances utiles et interprétables constitue donc un challenge scientifique et technologique majeur à la croisée de plusieurs domaines de recherche : bases de données, fouille de données, apprentissage automatique, psychologie cognitive, etc. Un objectif commun à ces différentes recherches est d’aboutir à un système dit intelligent d’accès aux données. Après avoir introduit les verrous scientifiques et technologiques à lever pour atteindre cet objectif, j’exposerai ma conviction que la théorie des sous-ensembles flous peut jouer un rôle central dans le développement d’une chaîne intelligente de transformation de données en connaissances.