My approach consists in exploiting statistical properties, multiscale analysis, non-linear modelisation, the physical properties of the signals, in order to propose new ways to deal with multi-disciplinary problems. More generally, I seek to extract most information out of experimental data, how that information can be used to build better models and produce better predictions.
I currently work on a new modeling technique. How, starting from observations on a given physical process, to build a reliable model of that process dynamics? When multiple processes interact at different scales, how to obtain a significant model at each of these scales? The goal is to provide a model simple enough to bring some understanding of the system studied, but also a model elaborated enough to allow precise predictions. In order to do so, this project proposes to identify causally equivalent classes of system states, then model their evolution with a stochastic process. Applications primarily concern natural sciences.
This work extends some earlier results to the general case of continuous time, (nearly-)arbitrary data. In order to do so, I have obtained the recognition of Inria's « Exploratory Actions » program, which allows me to fully dedicate time to this fundamental research. I also collaborate with the Complexity Science center at UC Davis.
I have previously worked on remote sensing of the environment and on granular matter physics, domains in which I still maintain some collaborations. That work consisted of the analysis of experimental signals, modelling, designing new algorithms and running numerical simulations.
I am a tenured member of the Geostat team at INRIA Bordeaux.
I was previously affiliated with the following groups and projects:
- I worked at the Institute of Physics of Rennes and at the Physics department of Duke University.
- I was on the board of the French association on Artificial Intelligence (AFIA), for two years.
- I animated for a few years the Inter-disciplinary Research network on Complex Systems applied to the Environnement (RISC-E).