Research

My research lies at the interface between optimization, statistics and machine learning. I develop and analyze algorithms for learning problems, in order to explain their observed behavior and to provide actionable guidelines. My current work focuses on the long-run behavior of stochastic gradient methods and on distributionally robust learning.

Research topics

Long-run behavior of stochastic gradient methods

Most machine learning models are trained by stochastic gradient descent (SGD) or one of its variants. With a finite training budget, these methods do not exactly minimize the training loss. Instead, they output a random model whose probability distribution depends on the loss landscape and on the gradient noise. Using large deviations theory, we characterize this long-run distribution, as well as the time needed to reach a global minimizer in non-convex landscapes. We have recently extended this analysis to learning in games.

Distributionally robust optimization and learning

Wasserstein distributionally robust optimization (WDRO) trains models that perform well on all probability distributions close to the empirical one, in the sense of optimal transport. We study the statistical guarantees of this approach, notably exact generalization bounds. Regularization makes these problems tractable, and we have applied them to risk control in prediction and to shape optimization. To make these models easy to use in practice, we develop the Python library skwdro.

Optimal transport and privacy

Optimal transport maps are used to compare and align datasets. We study how to estimate these maps from samples under differential privacy constraints, and we characterize the minimax rates of this problem.

Structure in nonsmooth optimization

Nonsmooth regularizers, such as the ℓ1 norm, induce structure in the solutions of learning problems (sparsity, low rank). Proximal methods identify this structure after a finite number of iterations. My habilitation studies how to detect and exploit this structure, for instance to accelerate algorithms with Newton steps on the identified manifold or to reduce the communications of distributed methods.

Distributed and asynchronous optimization

When data and computations are spread over several machines, the agents must cooperate although communications are delayed, costly or intermittent. I have designed and analyzed asynchronous, randomized and delay-tolerant algorithms for this setting, from gossip and ADMM methods during my PhD to multi-agent online learning.

Variational inequalities and games

Adversarial training and learning in games lead to variational inequalities rather than minimization problems. We have analyzed extragradient and optimistic methods in stochastic settings, with adaptive step-sizes, as well as the convergence rates of Bregman proximal methods.

Team

Postdocs

Matthieu Massin
from 2027
Robust optimization and statistics
Rodrigo Maulén-Soto
since 2026
Bias in generative models
Co-supervised with J.-M. Loubes and J. Bolte

PhD students

Marie Boyer
since 2026
Bias amplification in statistical learning
Co-supervised with X. Mai (Univ. Toulouse Jean Jaurès)
Iyad Zekhnini
since 2026
Stochastic optimization for non-convex machine learning objectives
Co-supervised with J. Malick and T. Le (CNRS & LJK, Grenoble)

Former PhD students

Master's interns

Show the list
  • Marie Boyer,
  • Iyad Zekhnini, two internships
  • Florian Vincent (UGA), LJK, Grenoble
  • Julien Prando (UGA), LJK, Grenoble
  • Waïss Azizian (ENS Paris & MVA), LJK, Grenoble (three internships), co-supervised with J. Malick and P. Mertikopoulos
  • Gilles Bareilles (ENSTA & MVA), LJK, Grenoble
  • Yu-Guan Hsieh (ENS Paris & MVA), LJK, Grenoble, co-supervised with J. Malick and P. Mertikopoulos
  • Konstantin Mishchenko (MIPT & MASH), LJK, Grenoble, co-supervised with J. Malick and M.-R. Amini
  • Taha Essalih, LEME, Paris Nanterre, co-supervised with N. El Korso, A. Breloy and R. Flamary

Grants and projects

Ongoing

Past

Service

Responsibilities

Scientific community

Organization of events

Software

skwdro

Python library for Wasserstein distributionally robust machine learning, with PyTorch and scikit-learn interfaces.