Reminder : All About that seminar series - Challenges in high dimensional Bayesian modeling, November 7, 2025
All About That Bayes
English version below
Chères et chers collègues,
petit rappel concernant de la séance du séminaire « All About that… » qui aura lieu le 7 Novembre 2025 à l'IHP, Paris (Plan d'accès)!
L’après-midi est accessible à toutes et tous, mais merci d’indiquer votre participation en vous inscrivant via le lien suivant :
https://forms.office.com/e/nA01fULXqdEn espérant vous voir nombreuses et nombreux!
Bien cordialement,
Kaniav Kamary et Merlin Keller
Pour le Groupe Spécialisé « Statistique Bayésienne » de la Société Française de Statistique (SFdS)
--------------------------------------------------------------------------------------------------------
Dear colleagues,
Quick reminder about the next session of the "All About that..." seminar series that takes place on November 7, 2025 at IHP, Paris (Access map)!
The event is open to all, but please confirm your attendance by registering via the following link:
https://forms.office.com/e/nA01fULXqd
We look forward to seeing many of you there!
Best regards,
Kaniav Kamary and Merlin Keller
For the Bayesian Statistics Group of the Société Française de Statistique (SFdS)
Theme: Challenges in high dimensional Bayesian modeling (Enjeux de la modélisation bayésienne en grande dimension)
Time: From 2 p.m. (14h00) to 5 p.m. (17h00)
Location: Pierre Grisvard (formaly room 314), IHP (Institut Henri Poincaré - Sorbonne Université / CNRS)
Full Program:
14h00 - 15h00 Julyan Arbel (Inria, Université Grenoble Alpes, Website) - Bayesian deep learning, overview and challenges
Abstract: Bayesian deep learning is appealing as it combines the coherence and natural uncertainty quantification of the Bayesian paradigm together with the expressivity and compositional flexibility of deep neural networks. It has its roots in pioneering work by Radford Neal and David Mackay in the 1990s on Bayesian neural networks. Its strengths lie in principled uncertainty quantification, improved data efficiency, and adaptability, making it impactful in safety-critical fields like healthcare and autonomous systems. In this talk I will provide an overview of Bayesian deep learning and discuss some of the key challenges the field faces in addressing modern machine learning problems.
15h00 - 16h00 Marion Naveau (Institut Agro Rennes-Angers, Website) - High-dimensional variable selection in non-linear mixed effects models. Application in plant breeding
Abstract: The problem of variable selection in high-dimensional context, where the number of covariates exceeds the number of observations, is well studied in the context of standard regression models. However, few tools are currently available to address this issue for nonlinear mixed-effects models, where data are collected repeatedly across multiple individuals. My thesis focused on developing a high-dimensional variable selection procedure for these models, examining both its practical implementation and theoretical properties. This method is based on a Gaussian spike-and-slab prior and the SAEM algorithm (Stochastic Approximation of the Expectation-Maximization Algorithm). Its utility is illustrated through an application aimed at identifying genetic markers potentially involved in the senescence process of winter wheat. Furthermore, metamodeling approaches are being developed to reduce computation time when the regression function is costly to evaluate.
16h00 - 17h00 Nicolas Chopin (ENSAE, Institut Polytechnique de Paris, Website) - Saddlepoint Monte Carlo and its Application to Exact Ecological Inference
Abstract: Assuming X is a random vector and A a non-invertible matrix, one sometimes need to perform inference while only having access to samples of Y = AX. The corresponding likelihood is typically intractable. One may still be able to perform exact Bayesian inference using a pseudo-marginal sampler, but this requires an unbiased estimator of the intractable likelihood. We propose saddlepoint Monte Carlo, a method for obtaining an unbiased estimate of the density of Y with very low variance, for any model belonging to an exponential family. Our method relies on importance sampling of the characteristic function, with insights brought by the standard saddlepoint approximation scheme with exponential tilting. We show that saddlepoint Monte Carlo makes it possible to perform exact inference on particularly challenging problems and datasets. We focus on the ecological inference problem, where one observes only aggregates at a fine level. We present in particular a study of the carryover of votes between the two rounds of various French elections, using the finest available data (number of votes for each candidate in about 60,000 polling stations over most of the French territory). We show that existing, popular approximate methods for ecological inference can lead to substantial bias, which saddlepoint Monte Carlo is immune from. We also present original results for the 2024 legislative elections on political centre-to-left and left-to-centre conversion rates when the far-right is present in the second round. Finally, we discuss other exciting applications for saddlepoint Monte Carlo, such as dealing with aggregate data in privacy or inverse problems.