LaBRI/INRIA GRAVITÉ Team
Locations of visitors to this page
May 2013
<<     >>
Mon Tue Wed Thu Fri Sat Sun
  1 2 3 4 5
6 7 8 9 10 11 12
13 14 15 16 17 18 19
20 21 22 23 24 25 26
27 28 29 30 31  
Username

Password

Remember Me!
Welcome on GRAVITÉ's web site > Want to work with us ? > Post-doctoral position

Post-doctoral position

To apply on the program, either contact the project head or go to INRIA's website and search for a position in Bordeaux.

This position requires Skills

Graph (data) mining; graph drawing and visualisation; human-computer interaction; certified skills developing with C++.

The GRAVITÉ INRIA team members have international expertise in graph visualization. The Tulip graph visualization framework (www.tulip-software.org) developed by our team gathers most of our results and those of the InfoViz community. Tulip will serve as a basis for all development performed during the project. Prior knowledge of Tulip is a plus.



Visualizing Dynamic Graphs through Multilevel representations / Focus+context views / Zoomable Interfaces with Applications to Semantic Networks

This one-year project requires background knowledge on existing results and literature concerning " focus+context " visualization and zoomable interfaces -- which motivates us to hire at a post-doc level.

We also expect the selected candidate to contribute prototypes and development efforts with our visualization framework Tulip, and more generally within the TANGUY architecture (see the TANGUY project page).

The post-doc year should also be the occasion to engage a survey study in an effort to output a unifying view on on multi-level visualization, focus+context techniques and zoomable interfaces. Typically, multi-elevel visualization rely on a hierarchization of the information space, itself allowing a focus+context process to act at the data and/or view level.

The TANGUY project requires that these techniques be adapted to graph mining and more particularly to deal with dynamic graphs. This raises a number of issues:
  • How can the hierarchy adapt to evolution of the graph ? Should we consider a just-in-time strategy to compute the hierarchy only when necessary ? What statistics -- if any -- can be used to drive the computation of the hierarchy based on graph changes ?
  • Recall how focus+context views are built: the hierarchy often is captured as a cluster tree (nested subgraphs); a view then corresponds to a cut in the tree.

  • The question is to decide/predict how the cut is affected by the dynamic changes occuring in the graph structure.
The first question is linked with issues on graph clustering extended to dynamic graphs. The idea of a just-in-time strategy also requires to have a closer look at how neighborhoods are defined/computed for dynamic graphs. The hierrachy itself can be somewhat generalized by allowing clusters to overlap, and thus to extend it directed acyclic graphs.