Winterhalder Group · TIFLab · Milano
Machine learning for
high-energy physics.
We develop and build neural surrogates and generative models for simulation and data analysis in particle physics, while asking what these models are really learning about the underlying physics.
What we work on
Generative models & fast simulation
Neural samplers and generative networks that speed up event generation and Monte Carlo integration.
Neural surrogates & uncertainties
Fast, calibrated surrogates and ML-based fits — including parton distributions — with uncertainties that can be trusted.
Inverse problems & simulation-based inference
Unfolding, simulation-based inference, and extracting physics parameters when the likelihood is intractable.
Interpretability
Asking what trained networks actually learn — from jet taggers to parton distributions.
Recent publications
- 2026
- 2026
- 2026
- 2026
- 2025