Research interests
My research interests are related to mathematical optimization, in particular nonsmooth analysis/optimization. Nonsmoothness typically arises in a highly structured way (from e.g. duality, min-max, penalization...). I study the underlying geometry of nonsmooth optimization problems to get a better understanding of algorithms, or to design new ones. Machine/deep learning is a very nice source of such problems. I illustrate all this with my personal top-5 papers.
Research topics
- Optimization under uncertainty: stochastic and (distributionally) robust optim.
- Nonsmooth optimization: theory and algorithms
- Optimization for machine learning
Highlight: Robustify your model with skWDRO
You have a learning model in pytorch, too sensitive to poor data or training/testing shifts ? Use our library SkWDRO to robustify your model ! You can train your model with (regularized) Wasserstein robustness, with just 3 lines of code ! Check out skWDRO github and the online doc, for practical examples and further information.
Highlight: good conferences...
With a good group, we have welcomed, one year out of two, in a ski resort close to Mont Blanc, some of the leading researchers in optimization and learning, for the serie of workshops OSL: we had a great fifth (and last) edition in January 2023 OSL2023. The serie has adapted and evolved to become ``Physics of AI algorithms'' with new orgnizers: check the nice 2025 session. Good luck for this nice follow-up.