Big Data Seminar

Le jeudi de 11 à 12h au LaBRI (salle 178)

Le séminaire Données massives et hétérogènes est ouvert à tous. Il traite de:

  • analyse/fouille de données/visualisation
  • convergence HPC/Big Data
  • algorithmique distribué/approximé/de graphes
  • IA (apprentissage, deep learning, ...)

L'objectif est de faire une programmation régulière et des présentations assez vulgarisées afin que tous puissent suivre.

Pour vous inscrire:

  • envoyer un mail à l'adresse
  • et mettre comme titre subscribe <prenom> <nom>

PROGRAMME (prévisionnel):

  • 13 septembre 2018: Automating Machine Learning

par André C. P. L. F. de Carvalho (Sao Paulo University)

As the number of successful applications of Machine Learning algorithms grows, there is also an increase in the need to make these algorithms easily accessible by users without Machine Learning expertise. There have been several efforts in this direction, involving not only the recommendation of the most suitable algorithm, but also their most appropriate hyper-parameter values. These several efforts started a new research area, named Automated Machine Learning, AutoML, which has attracted the attention of researchers and practitioners not only from the academia, but also from several companies working with data science. This talk will present the main approaches and recent advances in this area, covering also works carried out in the Analytics Laboratory, at USP São Carlos.

  • 7 juin 2018: Data Management in the Cloud: Evolution or Crossroad? par Abdelkader Hameurlain (IRIT, Toulouse)

RESUME: In the landscape of database management systems, data analysis systems (OLAP) and transaction processing systems (OLTP) are separately managed. The reasons for this dichotomy are that both systems have very different functionalities, characteristics and requirements. My talk will focus on the first class of OLAP systems. The purpose of this talk is twofold: (i) to provide a synthetic state of the art concerning (large-scale) data management systems, and (ii) how can the evolution of these systems help for big data applications. In this perspective, data management based on parallel and cloud systems are overviewed and compared by relying on fundamental criterion such as Software Requirements (Data Independence, Software Reuse), High Performance, Data Availability, Scalability and Elasticity . With respect to the state of the art, proposed systems, and qualitative and quantitative comparative studies between Parallel DBMS (PDBMS) and Big Data Management Systems (BDMS), we try to learn some lessons and point out some open issues that should be tackled to ensure the viability of the next generation of large-scale data management systems for big data applications.

Key Words: Big Data Management, Data Partitioning, Data Integration, Parallel Database Systems, Cloud Data Management Systems, Query Processing and Optimization, High Performance, Scalability, Elasticity, Hadoop MapReduce, Spark, Multistore Systems.

  • 24 mai 2018 Réseaux multi-couches, sciences humaines et sociales et bases de données de graphes : la nécessité plutôt que l’obsession du “big data" (suite et fin)
  • 28 avril 2018: Réseaux multi-couches, sciences humaines et sociales et bases de données de graphes : la nécessité plutôt que l’obsession du “big data" par Bruno Pinaud (LaBRI)
  • 22 mars 2018: Quand les Neurosciences Cognitives revisitent l'Intelligence Artificielle (partie 2) par Frédéric Alexandre (INRIA/LaBRI)
  • 8 mars 2018: Albert Biffet (Telecom Paris Tech) Massive Online Analytics for the Internet of Things

Big Data and the Internet of Things (IoT) have the potential to fundamentally shift the way we interact with our surroundings. The challenge of deriving insights from the Internet of Things (IoT) has been recognized as one of the most exciting and key opportunities for both academia and industry. Advanced analysis of big data streams from sensors and devices is bound to become a key area of data mining research as the number of applications requiring such processing increases. Dealing with the evolution over time of such data streams, i.e., with concepts that drift or change completely, is one of the core issues in stream mining. In this talk, I will present an overview of data stream mining, and I will introduce some popular open source tools for data stream mining.

  • 1 mars 2018: Les temps modernes de l'intelligence artificielle (partie 2) par Laurent Simon (LaBRI)
  • 8 février 2018: Quand les Neurosciences Cognitives revisitent l'Intelligence Artificielle par Frédéric Alexandre (INRIA/LaBRI)
  • 1er février 2018: Présentation de LSD par David Auber
  • 25 janvier 2018: Présentation de PlaFRIM par François Rue
  • 18 janvier 2018: Les temps modernes de l'intelligence artificielle par Laurent Simon (LaBRI)
  • 11 janvier 2018: Vers une convergence Calcul Haute Performance et Données Massives (suite), par Olivier Beaumont (INRIA, LaBRI)
  • 21 décembre 2017: Vers une convergence Calcul Haute Performance et Données Massives, par Olivier Beaumont (INRIA, LaBRI)
  • 7 décembre 2017 ( à 10h ),  Multiplication of Networks  par Vladimir Batagelj (University of Ljubljana).

Un exposé qui promet des solutions au problème posé par la forte densité des produits de réseaux.

  • 30 novembre 2017: Validation en Streaming (Linear Time Membership in a Class of Regular Expressions with Counting, Interleaving and Unordered Concatenation), par Dario Colazzo (Lamsade - Paris Dauphine)

Regular Expressions (REs) are ubiquitous in database and programming languages. While many applications make use of REs extended with interleaving (shuffle) and unordered concatenation operators, this extension badly affects the complexity of basic operations, and, especially, makes membership checking NP-hard, which is unacceptable in most practical scenarios.

In this talk I will present results about recent work where we study the problem of membership checking for a restricted class of these extended REs, called conflict-free REs, which are expressive enough to cover a wide class of real-world applications. Several polynomial algorithms for membership checking over conflict-free REs will be presented. The algorithms are all polynomial and differ in terms of adopted optimization techniques, ensuring linearity, and in the kind of supported operators. As a particular application, the approach can be generalized in order to efficiently check membership of XML trees into a class of Extended DTDs which models the crucial aspects of DTDs and XSD schemas

  • 23 novembre 2017: Introduction au deep learning, par Vincent Lepetit (LaBRI)
  • 6 avril: Résumé de données dans les flux par Olivier Beaumont ou Y. B.
  • 30 mars: Calcul distribué de clefs par Sofian Maabout
  • 23 mars: Star PU et ordonnancement dynamique par Samuel Thibault
  • 16 mars: Présentation des plates-formes de calcul PLAFRIM et LSD, David Auber et Olivier Coulaud ?
  • 9 mars (salle 178): Exploration de grands graphes et de cartes de chaleur par David Auber et Alexandre Perrot

La démocratisation du calcul distribué au travers de système tels que Hadoop ou Spark permet d'envisager de traiter de grands jeux de données en des temps raisonnables. Nous montrerons comment il est possible d'utiliser l'architecture Big Data pour effectuer un précalcul permettant de visualiser de grands jeux de données en adaptant des algorithmes existants. Nous utiliserons deux exemples : les cartes de chaleur et la visualisation de graphe

  • ?? (salle 178): Convergence HPC/Big Data et analyse de données issus de capteurs par Olivier Beaumont et Nicolas Hanusse
  • 14 février (batiment INRIA): Modélisation de l'architecture d'un super-ordinateur par Guillaume Aupy
  • 9 février (Amphi LaBRI): Data Science: opportunities and risks par Patrick Valduriez (INRIA Sophia-Antipolis, LIRMM)

Data has been quoted as the new oil, to reflect that big data can be turned into high-value information and new knowledge. Although data analysis has been around for a while, starting with statistics and evolving lately into exploratory data analysis, data mining and business intelligence, the new dimensions of big data (volume, variety, velocity, etc.) make it very hard to process and analyze data, and derive good conclusions. To address this grand challenge, data science is emerging as a new science that combines computer science, statistics and machine learning, visualization and human-computer interactions to collect, clean, integrate, analyze and visualize big data. The ultimate goal is to create new data products and services, as well as training legions of data scientists. In this talk, I will introduce data science, including big data and cloud technologies. I will illustrate the main opportunities and risks, in particular by telling my favorite stories about the good, the bad and the ugly. Finally, I will discuss the new challenges in combining HPC and big data analytics.

  • 2 février (salle 178): présentation de l'axe - discussion/organisation/planning par Nicolas Hanusse (LaBRI)

En relation avec le séminaire

  • 10 mars 14h (Amphi LaBRI): Optimisation de requêtes sur des données massives dans un environnement distribué, soutenance de thèse de Noël Gillet
  • 9 février 14h (Amphi LaBRI): Optimisation de la requête multi-critères SKyline en fouille de données, soutenance de thèse de Patrick Kamnang-Wanko.
  • 6 février, 11h (Amphi LaBRI): '''Présentation des projets de Mappy: calcul de routes et vues immersives"