User Tools

Site Tools



The SPARKS team is composed by 3 Themes.

Knowledge Extraction and Learning

The activity in this research theme has mainly focused on the development of methods and algorithms exploiting ideas and techniques from machine learning, data mining, natural language processing, and artificial intelligence for extracting novel, useful information and knowledge from data. Many kinds of data have been considered, ranging from structured data in tabular form (records with numerical and/or categorical attributes) to unstructured data such as text in natural language, graphs (as diverse as social networks, see also FORUM theme, gene regulatory networks, and linked data) and multimedia data, including time series (e.g., electrocardiographic signals), images, videos and 3D data. The methods and techniques used include support-vector machines, neural networks, boosting, decision trees, random forests, frequent pattern and association rule mining , inductive logic programming, fuzzy set theory, and evolutionary algorithms.

A particular emphasis has been given to the scalability of the approaches, according to three dimensions:

  1. the volume of the data that are to be processed;
  2. the number of processing units available for computation and the distributed nature of both data and algorithms;
  3. the computational power of the processing units, which may become critical when the proposed approaches must be embedded in appliances, vehicles, or mobile devices, often with tight resource consumption constraints.

FOrmalizing and Reasoning with Users and Models (FORUM)

Activities within this research theme address the general problem of reconciling formal semantics of computer science (e.g. logics, ontologies, typing systems, protocols, etc.) on which the Web architecture is built, with soft semantics of people (e.g. posts, tags, status, relationships, etc.) on which the Web content is built. The research works in this theme contribute to the understanding of these graphs by (1) proposing multidisciplinary approaches to analyze and model the many aspects of these intertwined information systems, their communities of users and their interactions, and (2) formalizing and reasoning on these models using graph-based knowledge representation from the semantic Web to propose new analysis tools and indicators, and support new functionalities and better management.

The research objectives of this theme can be grouped according to three main topics: (1) supporting Human Computer Interaction (HCI) and Human Data Interaction (HDI), (2) supporting interaction between users and (3) reasoning and interacting with knowledge graphs. As a corollary, we design user-centered models and methods and we conduct user experimentations and user evaluations of the algorithms and platforms we develop.

Scalable Software Systems

With the generalisation of multi-cores and distributed computing resources, programmers face new software parallelisation, adaptability, and scaling-up complexities. SPARKS research in that field is grounded on a long-standing experience in the domains of scientific workflow systems that facilitate distributed computing infrastructures exploitation by non-expert users, scalable and secure software composition techniques, and adaptation to heterogeneous and dynamic execution environments. In addition, part of the activity complements the research theme FORUM theme with concerns related to large-scale distributed data repositories integration.

themes.txt · Last modified: 2016/10/11 11:23 by philippereneviergonin