Beyond Bayesian Uncertainty: a Variational Perspective
In this talk, I will explore how the statistics and machine learning communities are expanding the frontiers of uncertainty quantification beyond traditional Bayesian frameworks.
Overview
I will discuss the motivations behind this shift and introduce emerging variational approaches that offer fresh perspectives on inference. Drawing on my own research, I will highlight key milestones in this evolving landscape and outline some of the remaining challenges. Finally, I will show how variational methods open new avenues, overcoming limitations where classical Bayesian approaches fall short.
Presenters
Chérief-Abdellatif Badr-Eddine, CNRS (Chargé de Recherche) Researcher, Laboratoire de Probabilités, Statistique et Modélisation (LPSM), Sorbonne Université, France
Brief Biography
Chérief-Abdellatif Badr-Eddine is a CNRS researcher (Chargé de Recherche) based at Sorbonne Université, in the Laboratoire de Probabilités, Statistique et Modélisation (LPSM), in Paris. Prior to that, he was a postdoctoral researcher in the Department of Statistics at the University of Oxford, and he received his Ph.D. in 2020 from Institut Polytechnique de Paris, prepared in the Center for Research in Economics and Statistics (CREST), under the supervision of Professor Pierre Alquier. His research spans several areas of mathematical statistics and machine learning theory, with a particular focus on generalized Bayesian inference. More broadly, he is interested in variational inference, PAC-Bayesian theory, robustness, generalization bounds, kernel methods, and missing data analysis.