About me
My scientific interests lie at the interface between mathematics, physics and computer science.
After a curriculum in mathematics and physics, I notably studied differential geometry, gauge theory and operator algebra, mostly motivated by the challenges posed by quantum physics and general relativity.
My doctoral research then focused on statistical physics and graphical models. I used algebraic topology to describe Bethe-Kikuchi variational principles and generalised belief propagation (GBP). I proposed new regularisations of this message-passing algorithm, which efficiently estimate the marginals of a high dimensional probability distribution, and take the form of continuous-time diffusion equations.
I am now working as a postdoc at the Computer science Research Institute of Lens (CRIL). I focus on packaging an efficient implementation of graphical models and message-passing algorithms for the GPU. I also wish to extend the discrete algebraic and geometric framework developed during my PhD to wider applications, such as continuous or quantum graphical models.
Publications
- Local Max-Entropy and Free Energy Principles solved by Belief Propagation, 2023 (short preprint). Presented at the MaxEnt 2022 conference, extended article awaiting publication in Entropy.
- Belief Propagation as Diffusion, 2021, Geometric Science of Information, Frederic Barbaresco and Frank Nielsen editors, Springer (proceedings of GSI 21).
- Message-Passing Algorithms and Homology: from thermodynamics to statistical learning, 2020, Thèses, Université Paris Cité.
- Homological Algebra for Message Passing Algorithms, 2020, presented at the SYCO 2020 conference in Tallinn (preprint).
- A Homological Approach to Belief Propagation and Bethe Approximations, 2019, Geometric Science of Information, Frederic Barbaresco and Frank Nielsen editors, Springer (proceedings of GSI 19).
as coauthor
- Equivariant Message Passing Neural Networks for Material Discovery, 2023, AAAI (preprint) – with Astrid Klipfel, Yaël Frégier, Zied Bouraoui and Adlane Sayede.
- Extrafine Sheaves and Interaction Decompositions, 2023, under revision for Algebraic and Geometric Topology (preprint) – with Daniel Bennequin, Grégoire Sergeant-Perthuis and Juan-Pablo Vigneaux.
- Geomstats: a Python Package for Riemannian Geometry in Machine Learning, 2020, Journal of Machine Learning Research, 21 (223), pages 1-9 – with Nina Miolane, Nicolas Guigui, Alice Le Brigant, Johan Mathe, Benjamin Hou, Yann Thanwerdas, Stefan Heyder, Niklas Koep, Hadi Zaatiti, Hatem Hajri, Yann Cabanes, Thomas Gerald, Paul Chauchat, Christian Shewmake, Daniel Brooks, Bernhard Kainz, Claire Donnat, Susan Holmes and Xavier Pennec.
- Introduction to Geometric Learning in Python with Geomstats, 2020, SciPy 2020 - 19th Python in Science Conference – with Nina Miolane, Nicolas Guigui, Hadi Zaatiti, Christian Shewmake, Hatem Hajri, Daniel Brooks, Alice Le Brigant, Johan Mathe, Benjamin Hou, Yann Thanwerdas, Stefan Heyder, Niklas Koep, Yann Cabanes, Thomas Gerald, Paul Chauchat, Bernhard Kainz, Claire Donnat, Susan Holmes and Xavier Pennec.