Research
While machine learning systems can be incredibly powerful, it is of crucial interest to understand how confident we can be in their outputs.
My work focuses on providing statistical guarantees for these predictions.
In particular, I do research on classifier calibration, proper scoring rules, and conformal prediction.
|
Publications
|
Multivariate Conformal Prediction via Conformalized Gaussian Scoring
Sacha Braun,
Eugène Berta,
Michael I. Jordan,
Francis Bach
arXiv preprint, 2025
arxiv /
code
|
Rethinking Early Stopping: Refine, Then Calibrate
Eugène Berta,
David Holzmüller,
Michael I. Jordan,
Francis Bach
arXiv preprint, 2025
arxiv /
code /
slides
|
Classifier Calibration with ROC-Regularized Isotonic Regression
Eugène Berta,
Francis Bach,
Michael I. Jordan
Proceedings of the International Conference on Artificial Intelligence and Statistics (AISTATS), 2024
paper /
code
|
Software
I am developping the probmetrics Python package with David Holzmüller.
The package provides efficient implementations for widely used post-hoc calibration functions like Temperature Scaling.
|
Teaching
Starting from September 2025, I'll be a teaching assistant for the second year course "Mathématiques fondamentales" in the "Cycle Pluridisciplinaire d’Études Supérieures" (CPES) of Lycée Henri IV and PSL university.
|
|