PhD position in Visual Analytics in Marseille, South of France

Incremental computation and interactive visualization of dependencies and anomalies discovery in extremely large databases.

Project Description:
In databases of extreme size from terabytes to petabytes of data, the data exploration and discovery of interesting patterns, dependencies or anomalies is extremely challenging. Current data mining methods (e.g. for dependency inference, TANE, FUN, FastFDs) may be efficient for small to middle-size databases as long as the dataset can fit in memory but, for extra-large databases, they fall short and are completely inoperative due to memory overload.

The interactive exploration of such datasets requires incremental and visual computing with a careful management of memory transfers and the repartition of the computations on-the-fly on several processing units (GPU, CPU). Additionally, advanced artifacts of visual analytics should be offered to the users, so they can easily understand the relationships discovered from the data and interactively refine the grain of the analysis.

The successful candidate will examine these requirements and propose new incremental and visual mining algorithms and develop a visualization tool to test on extra-large scientific datasets in astronomy and environmental sciences.