I am a postdoctoral researcher in the PML group at Aalto University (Finland), working with Pr. Samuel Kaski. I used to work at Télécom ParisTech in the DIG team (former DBWeb), where I did my PhD under the supervision of Jean-Louis Dessalles and Antoine Cornuéjols. My research interests are learning theory and especially transfer learning. I work with the tools of algorithmic information theory (Kolmogorov complexity and Minimum Description Length Principle) and I am convinced that these tools might be a valuable alternative to statistical modeling when it comes to understanding why and when machine learning works.
I have also worked on a couple of thrilling side projects:
Collaborative clustering: The idea is to exchange information between several clustering algorithms in order to make them refine their decisions. An interesting question I investigate: What are the guarantees that collaborating is of any interest for the algorithms?
Analogical reasoning: How to evaluate the quality of statements of the form "A is to B what C is to D"?
Geometrical modeling of learning: Is it possible to represent the non-stationary of data by a riemannian curvature?