I am a research scientist at Inria, in Sophia Antipolis, near Nice. I am part of the Maasai (Models and Algorithms for Artificial Intelligence) team. I am also affiliated with the J.A. Dieudonné lab, which is the mathematics research department of Université Côte d'Azur. I hold a chair of the 3IA Côte d'Azur.

I received in 2017 my Ph.D. in applied mathematics from Université Paris Descartes (now called Université Paris Cité). I worked at the MAP5 lab, where I had the chance to be advised by Charles Bouveyron and Pierre Latouche. After my Ph.D., I did a postdoc at the IT University of Copenhagen, where I mainly worked with Jes Frellsen. My field of research is statistical machine learning, with a particular emphasis on hidden variables and model uncertainty.

I am co-organising at teaching at the Generative Modeling Summer School that was held in Copenhagen in 2023, and will be held in Eindhoven in 2024. Applications are open for 2024!

I am also one of the co-organisers of the Workshop on the Art of Learning with Missing Values (Artemiss). You can watch videos of the first edition (that was part of the ICML 2020 conference) here. I am also co-organising two annual workshops: the Workshop on Generative Models and Uncertainty Quantification (GenU), a small-scale workshop generally held in October in Copenhagen, and Statlearn, a workshop on statistical learning held in France every year.

Current PhD students and postdocs

Alumni

Teaching

Ecole Normale Supérieure de Paris Saclay
GeMSS summer school
Generative Modeling Summer School, Copenhagen 2023, Eindhoven 2024
Université Côte d'Azur

Preprints & working papers

Are ensembles getting better all the time?
(with Damien Garreau)
Preprint arXiv:2311.17885, 2023
Generalised Mutual Information: a Framework for Discriminative Clustering
(with Louis Ohl, Charles Bouveyron, Warith Harchaoui, Mickaël Leclercq, Arnaud Droit, and Frédéric Precioso)
Extended version of our NeurIPS 2022 paper, arXiv:2309.02858, 2023
Sparse GEMINI for Joint Discriminative Clustering and Feature Selection
(with Louis Ohl, Charles Bouveyron, Mickaël Leclercq, Arnaud Droit, and Frédéric Precioso)
Preprint arXiv:2302.03391, 2023
Fed-MIWAE: Federated Imputation of Incomplete Data via Deep Generative Models
(with Irene Balelli, Aude Sportisse, Francesco Cremonesi, and Marco Lorenzi)
Preprint arXiv:2304.08054, 2023
A Multi-stage deep architecture for summary generation of soccer videos
(with Melissa Sanabria, Frédéric Precioso, and Thomas Menguy)
Preprint arXiv:2205.00694, 2022
Uphill Roads to Variational Tightness: Monotonicity and Monte Carlo Objectives
(with Jes Frellsen)
Preprint arXiv:2201.10989, 2022
A Parsimonious Tour of Bayesian Model Uncertainty
Preprint arXiv:1902.05539, 2020

Publications

2023

Explainability as statistical inference
(with Hugo Senetaire, Damien Garreau, and Jes Frellsen)
Proceedings of the 40th International Conference on Machine Learning, pp. 30584-30612, 2023
Are labels informative in semi-supervised learning? Estimating and leveraging the missing-data mechanism
(with Aude Sportisse, Hugo Schmutz, Olivier Humbert, and Charles Bouveyron)
Proceedings of the 40th International Conference on Machine Learning, pp. 32521-32539, 2023
Don't fear the unlabelled: safe deep semi-supervised learning via simple debiasing
(with Hugo Schmutz and Olivier Humbert)
International Conference on Learning Representations, 2023

2022

Generalised Mutual Information for Discriminative Clustering
(with Louis Ohl, Charles Bouveyron, Warith Harchaoui, Mickaël Leclercq, Arnaud Droit, and Frédéric Precioso)
Advances in Neural Information Processing Systems, 2022
Asteroid Taxonomy from Cluster Analysis of Spectrometry and Albedo
(with Max Mahlke and Benoit Carry)
Astronomy & Astrophysics, vol. 665 (A26), 2022
Introduction à l'intelligence artificielle et aux modèles génératifs
(with Serena Villata)
Chapter in Informatique mathématique: Une photographie en 2022 (CNRS editions), edited by Bruno Martin and Sara Riva
Model-agnostic out-of-distribution detection using combined statistical tests
(with Federico Bergamin, Jakob Havtorn, Hugo Senetaire, Hugo Schmutz, Lars Maaløe, Søren Hauberg, and Jes Frellsen)
Proceedings of the 25th International Conference on Artificial Intelligence and Statistics, pp. 10753-10776, 2022
How to deal with missing data in supervised deep learning?
(with Niels Bruun Ipsen and Jes Frellsen)
International Conference on Learning Representations, 2022
Tensor decomposition for learning Gaussian mixtures from moments
(with Rima Khouja and Bernard Mourrain)
Journal of Symbolic Computation, vol. 113, pp. 193-210, 2022
Unobserved classes and extra variables in high-dimensional discriminant analysis
(with Michael Fop, Charles Bouveyron, and Brendan Murphy)
Advances in Data Analysis and Classification, vol. 16, pp. 55-92, 2022

2021

not-MIWAE: Deep Generative Modelling with Missing not at Random Data
(with Niels Bruun Ipsen and Jes Frellsen)
International Conference on Learning Representations, 2021

2020

Exact Dimensionality Selection for Bayesian PCA
(with Charles Bouveyron and Pierre Latouche)
Scandinavian Journal of Statistics, vol. 47 (1), pp. 196-211, 2020

2019

MIWAE: Deep Generative Modelling and Imputation of Incomplete Data Sets
(with Jes Frellsen)
Proceedings of the 36th International Conference on Machine Learning, pp. 4413-4423, 2019
Partially Exchangeable Networks and Architectures for Learning Summary Statistics in Approximate Bayesian Computation
(with Samuel Wiqvist, Umberto Picchini, and Jes Frellsen)
Proceedings of the 36th International Conference on Machine Learning, pp. 6798-6807, 2019
Class-specific Variable Selection in High-Dimensional Discriminant Analysis through Bayesian Sparsity
(with Fanny Orlhac, Charles Bouveyron, and Nicholas Ayache)
Journal of Chemometrics, vol. 33 (2), e3097, 2019

2018

Leveraging the Exact Likelihood of Deep Latent Variable Models
(with Jes Frellsen)
Advances in Neural Information Processing Signals 31, pp. 3859-3870, 2018
Refit your Encoder when New Data Comes by
(with Jes Frellsen)
3rd NeurIPS workshop on Bayesian Deep Learning, 2018
Bayesian Variable Selection for Globally Sparse Probabilistic PCA
(with Charles Bouveyron and Pierre Latouche)
Electronic Journal of Statistics, vol. 12 (2), pp. 3036-3070, 2018

2017

Model Selection for Sparse High-Dimensional Learning
Ph.D. Thesis, Université Paris Descartes, 2017
Multiplying a Gaussian Matrix by a Gaussian Vector
Statistics & Probability Letters, vol. 128, pp. 67–70, 2017
Discussion on the Paper "A Bayesian Information Criterion for Singular Models" by Drton and Plummer
Journal of the Royal Statistical Society: Series B, vol. 79, pp. 370–371, 2017

2016

Combining a Relaxed EM Algorithm with Occam's Razor for Bayesian Variable Selection in High-Dimensional Regression
(with Pierre Latouche, Charles Bouveyron, and Julien Chiquet)
Journal of Multivariate Analysis, vol. 146, pp. 177–190, 2016