Our work “Towards Understanding Steering Strength” with Magamed and Damien was accepted to ICML 2026!
News
With the ICML 2026 deadline, I forgot to mention that together with Sofiane Tanji and Yassine Laguel, we have a new preprint Fairness-informed Pareto Optimization: An Efficient Bilevel Framework. If you like fairness or multiobjective optimization or bilevel optimization, you won’t be disappointed to skim it!
A small Rust tool to monitor new email using IMAP IDLE command svaiter/imapidle_then
NeurIPS 2025 San Diego. I will be in San Diego for NeurIPS (December 2-6).
PhD defense: Léo Davy. Léo Davy defended his PhD thesis (supervisors: Nelly Pustelnik and Patrice Abry) that I had the pleasure to review. Congrats!
Talk at Séminaire Math-IA Bordeaux. I gave a talk at the Séminaire Math-IA in Bordeaux on Geometrical and computational complexity of bilevel optimization.
Promoted Directeur de recherche. Starting today, I am officially promoted Directeur de recherche at CNRS.
CIRM workshop on inverse problems. I am the CIRM this week for the (Blind) inverse problems in imaging: from foundations to applications workshop organized by Luca, Emmanuel and Pierre.
New PhD student: Tùng Nguyễn Thanh. Tùng Nguyễn Thanh just started his PhD with Yassine Laguel and myself.
Workshop at Simons Institute Berkeley. I am in Berkeley this week for the Graph Learning Meets Theoretical Computer Science workshop at the Simons Institute for the Theory of Computing.
Visiting Rennes. I am visiting Nicolas Keriven in Rennes this week.
PhD defense: Sophie Jaffard. Sophie Jaffard defended his PhD thesis on Spiking neural networks : learning as point processes. Congrats to her! It was a pleasure to supervise her with Patricia Reynaud-Bouret.
Café Calcul presentation. I present at Café calcul from Groupe Calcul. I introduce automatic differentiation: from the chain rule and Jacobians to forward and reverse modes (backprop) via computational graphs, showing simple implementations, semantic interpretations, and a glimpse of theory for iterative algorithms.
Talk at LJAD-MAASAI seminar. I gave a talk at the joint seminar between the Probability and Statistics team of the LJAD and the Inria MASAAI team on some advances on nonconvex bilevel optimization.
Talk at MaLGa Genoa. I gave a talk at MaLGa in Genoa, hosted by L. Calatroni.
Bilevel paper accepted to Math. Prog.. Our paper Geometric and computational hardness of bilevel programming has been accepted to Math. Prog!
Talk at GDR IASIS Lyon. I give a talk at GDR IASIS day on bilevel optimization in Lyon.
Oberwolfach workshop. I was in Oberwolfach for the Mathematical Imaging and Surface Processing workshop. The report is available.
Winter school Dijon-Besançon. I gave a 5-hours introduction to language models at the Winter school Dijon-Besaçon.
MIA 2025 Conference. We organized MIA 2025 Conference in Paris.
ICSDS 2024 Conference. We organized the ICSDS 2024 Conference in Nice.
NeurIPS 2024 Vancouver. I will be in Vancouver for NeurIPS.
PhD defense: Matthieu Cordonnier. Matthieu Cordonnier defended his PhD thesis on “Neural networks on large scale random graphs”. Congrats!
Talk at SophIA Summit. I gave an invited talk at SophIA Summit
MPGNN paper accepted to JMLR. Our paper Convergence of Message Passing Graph Neural Networks with Generic Aggregation On Large Random Graphs was accepted for publication in J. Mach. Learn. Res.
PhD defense: Mathieu Dagréou. Mathieu Dagréou defended his PhD thesis on “Contributions to stochastic bilevel optimization”. Congratulations Mathieu!
Derivatives of SGD accepted to NeurIPS 2024. Our (w/ F. Iutzeler & E. Pauwels) paper Derivatives of Stochastic Gradient Descent was accepted to NeurIPS 2024.
3IA chair holder. Just learned that I will be a 3IA chair holder for the next 4 years for my project BOGL (Bilevel Optimization for Graph Learning).
LOCA workshop. We organized with L. Calatroni the LOCA workshop in Sophie-Antipolis.
Talk at SPARNP conference. I gave a talk at the Statistical and Probabilistic Analysis of Random Networks and Processes conference.
One Piece of Math a Day. Started my tweet serie “One Piece of Math a Day” on Twitter. Also available on my website here.
Talk at Inria Grenoble. I gave a talk at Inria Grenoble on automatic differentiation.
Talk at EURO 2024. Attending EURO 2024. I will give an invited talk in the “Algorithms for machine learning and inverse problems” session.
ANR MAD grant accepted. The grant ANR MAD was just accepted. I will be the PI of the project “Mathematics of Automatic Differentiation” for the next 4 years with the following consortium: J. Bolte, S. Gadat, E. Pauwels (Toulouse School of Economics), G. Fort, F. Iutzeler (Université Paul Sabatier) and J.-B. Caillau, Y. Laguel (Université Côte d’Azur).
Visiting MILA. Visiting G. Gidel at MILA. I gave a talk at MTL Opt.
Gradient Scarcity accepted to TMLR. Accepted paper to TMLR with Hashem Ghanem and Nicolas Keriven Gradient Scarcity with Bilevel Optimization for Graph Learning with Featured Certification!
Preprint: CHANI. New preprint with Sophie Jaffard and Patricia Reynaud-Bouret CHANI: Correlation-based Hawkes Aggregation of Neurons with bio-Inspiration.
Preprint: Derivatives of SGD. New preprint with Franck Iutzeler and Edouard Pauwels on the Derivatives of Stochastic Gradient Descent.
HVP blogpost spotlight at ICLR 2024. Our blogpost with Mathieu Dagréou, Pierre Ablin and Thomas Moreau How to compute Hessian-vector products? was accepted to ICLR 2024 as spotlight.
Visiting Flatiron Institute NYC. Visiting A. Bietti at CCM, Flatiron Institute, NYC. I presented our work on iterative automatic differentiation.
Talk at PEPR PDE-AI kickoff. Talk at the kickoff meeting of the PEPR PDE-AI project.
Two papers accepted to AISTATS 2024. Our papers Provable local learning rule by expert aggregation for a Hawkes network and A Lower Bound and a Near-Optimal Algorithm for Bilevel Empirical Risk Minimization were accepted to AISTATS 2024.
Graph-Slope paper accepted in EJS. Our paper with Pierre C. Bellec and Joseph Salmon coined A sharp oracle inequality for Graph-Slope has been accepted in Electronic Journal of Statistics.
Preprint on Graph-Slope. Together with Pierre C. Bellec and Joseph Salmon, we released our preprint A sharp oracle inequality for Graph-Slope.
Model Consistency of Partly Smooth Regularizers accepted in TIT. After two long years in review, our paper with J. Fadili and G. Peyré Model Consistency of Partly Smooth Regularizers has been accepted for publication in IEEE Transactions on Information Theory.
Accelerated alternating descent methods published in JMIV. Our paper with A. Chambolle and P. Tan, Accelerated Alternating Descent Methods for Dykstra-like problems has been accepted for publication in Journal of Mathematical Imaging and Vision. The online first version is available here.
New preprint on non-uniqueness for the analysis Lasso. Our work with A. Barbara and A. Jourani on non-uniqueness for the analysis Lasso is available at the following link Maximal solutions of sparse analysis regularization.
CLEAR accepted in SIAM IS. Our paper on re-fitting has been accepted for publication in SIAM Journal on Imaging Science. The preprint is available at the following link CLEAR: Covariant LEAst-square Re-fitting with applications to image restoration.
New preprint on accelerated alternating descent methods. Together with A. Chambolle and P. Tan, we extended recent results by Antonin and T. Pock on the acceleration of alternating minimization techniques for quadratic plus nons-mooth objectives depending on two variables, with an emphasis on the strongly convex case. The paper is available at Accelerated Alternating Descent Methods for Dykstra-like problems.
New preprint on re-fitting. Together with C. Deledalle, N. Papadakis and J. Salmon, we proposed a new framework to remove parts of the systematic errors affecting popular restoration algorithms. This work is available at the following link CLEAR: Covariant LEAst-square Re-fitting with applications to image restoration.
Paper accepted. Our paper The Degrees of Freedom of Partly Smooth Regularizers has been accepted for publication in Annals of the Institute of Statistical Mathematics (AISM).
MOA Workshop 2015. We organize at the IMB the annual workshop of the GdR MOA in December 2015. It will highlight recent advances in optimization. More information here.
PhD PGMO award. I received one of the two PGMO PhD award for my thesis. This award is delivered by Programme Gaspard Monge pour l’optimisation et la recherche opérationnelle.
CNRS. My postdoctoral position is over and I am now a CNRS researcher (CR2). I joined the IMB at Université de Bourgogne.
PhD award. I received a best PhD award for my thesis. This award is delivered by Fondation Dauphine.
Two new papers. Our paper Model Selection with Low Complexity Priors has been accepted for publication Information and Inference (IMAIAI). Moreover, our book chapter Low Complexity Regularization of Linear Inverse Problems will be published in Sampling Theory, a Renaissance edited by G. Pfander.
New Position. I just started my postdoc with Antonin Chambolle at the CMAP, Ecole Polytechnique.
SUGAR. Our paper Stein Unbiased GrAdient estimator of the Risk (SUGAR) for multiple parameter selection has been accepted for publication in SIAM Journal on Imaging Sciences.
Ph. D.. I successfully defended my PhD dissertation, see this page for more information.
Preprints on Model Stability and Stein Estimation. Model Consistency of Partly Smooth Regularizers and Stein Unbiased GrAdient estimator of the Risk (SUGAR) for multiple parameter selection are now available online.
Preprint DOF. Our last preprint The Degrees of Freedom of Partly Smooth Regularizers is available online.