Romain BOURQUI Maître de Conférences |
Contact
bourqui [at] labri.fr LaBRI, Université de Bordeaux UMR CNRS 5800 351, cours de la Libération F-33405 Talence cedex France Tel.: +33 5 4000 3530 |
About me
Since 1st 2009, I am an Associate Professor in the Computer Science Department of the IUT ("Technical School"), University of Bordeaux (Talence), France. I am also head of the BKB ("Bench to Knowledge and Beyond") team of LaBRI. From 1st October 2008 to 31st August 2009, I was post-doctoral researcher in computer science with the VIS team of the Eindhoven University of Technology, Netherlands. My collaborators there were J. J. van Wijk and M. A. Westenberg. September 2005 to October 2008, Phd in Computer Science with LaBRI laboratory, University of Bordeaux: Research Activity ⬏
Topics
The development of data acquisition techniques makes possible the generation more and more data. The size and complexity of this data raises a new problem: do traditional data analysis tools (e.g. diagrams, spreadsheets, etc.) allow to explore them efficiently? It seems obvious that the analysis of a list of several thousand (or even millions or billions) of lines requires a significant effort for experts. This constant increase therefore makes crucial the design of data exploration and analysis tools that allow their exploitation. This is even the more true in the era of Big Data, where the size of data far exceeds the capabilities of conventional computers. The objective of my research work is therefore the development of such tools. In order to allow the visual exploration of complex and/or massive data, my work focuses on the definition of graph drawing algorithms and the definition of data summary visualizations, also called abstract visualizations. The improvement of computing power as well as the availability of large data sets, especially in distributed infrastructures, have allowed the development of supervised learning methods. However, these methods, and in particular deep neural networks, are difficult to explain. This makes them currently unusable in certain application domains such as health and security. These networks offer particularly interesting case studies due to their size and complexity. The current direction of my research is not only oriented towards visualization for supervised learning in order to explain the predictions, improve their performance or reduce their size in order to reduce the necessary computing resources (energy cost, implementation in connected objects), but also towards the use of these methods for the automatic evaluation of visualization techniques. Projects
On going project:
Previous projects:
Students supervision
On going supervision:
Former students:
Publications ⬏
Administrative ⬏
Research administration
Teaching administration
Teaching ⬏
Département Informatique, IUT de Bordeaux
Master 2 Bio-informatique, Université de Bordeaux
3ème année, parcours IA, ENSC/ENSEIRB-MATMECA
Software & prototypes ⬏
PIVERT: a suite of prototypes for interpreting Deep Neural Netowrks
PIVERT website can be found here.
rNAV: a visualization tool for bacterial sRNA-mediated regulatory networks mining
rNAV website can be found here.
Systrip: a visual environment for the investigation of time-series data in the context of metabolic networks
Systrip website can be found here.
MetaViz: Visualization software for metabolic network
MetaViz website can be found here.
MotusVis: Software for visual comparison of metabolic networks
A demo can be found here.
Tulip: Framework dedicated to the analysis and visualization of relational data
Tulip webstite can be found here.
|