Second PHYSTAT Topical Meeting on “Statistics meets ML” in Particle Physics & Astrophysics – VERaIPHY
In recent years, ML has become more and more integrated into many stages of our analyses. In Particle Physics, it includes data collection and processing (triggering, tracking, etc), classification of different particle types, unfolding, parameter determination, anomaly detection, and even end-to-end processing. In astronomy, its use is increasingly widespread in areas such as object classification, distance determination, and regression problems. It is used theoretically to enhance simulations and to emulate theoretical predictions that are expensive to compute. It is also increasingly used in simulation-based inference, both for finding informative summary statistics and for variational methods of inference.
The first PHYSTAT meeting on “Statistics meets Machine Learning” brought together experts from particle physics, astrophysics, statistics, and ML to discuss the statistical foundations and limitations of these approaches. While ML methods often outperform traditional techniques in terms of speed and apparent precision, a central open question remains whether they are also statistically reliable and scientifically trustworthy, in particular with respect to uncertainty quantification, generalization, robustness, and interpretability.
Since that first meeting, a coordinated community effort has emerged to systematically address these questions: the VERaIPHY initiative (“Validation & Evaluation for Robust AI in PHYsics”). Within VERaIPHY, ten focused working groups have been developing a series of in-depth review and research articles on the statistical validation of modern ML methods in the physical sciences.
The present meeting serves as a capstone event for this article series. Rather than a broad, open-call workshop, it is structured as a topical assessment and synthesis, in which the authors of the VERaIPHY working groups will present and discuss their results. The program therefore consists of invited contributions from these groups, followed by extended discussion sessions with the wider PHYSTAT community.
The scientific scope covers, among others:
-
Statistical properties of training and generalization
-
Uncertainty quantification and calibration
-
Model robustness and failure modes
-
Inference and parameter estimation
-
Generative models and their statistical validation
-
Hypothesis testing and discovery
-
Model selection, interpretability, and explainability
-
Information-theoretic aspects and data representations
-
Symbolic and physics-informed ML
Although the list of speakers is fixed, participation is open, and ample time will be reserved for questions and discussion with all attendees.
The meeting will be hybrid i.e. participation will be either in person or remote. In both cases, registration is necessary, as the Zoom link to the meeting will be sent only to registered participants a few days before the meeting.
In-person registration costs €50 to cover coffee and lunch. The limit for in-person participants is 80.
PHYSTAT
The PHYSTAT series of Workshops started in 2000. They were the first meetings devoted solely to the statistical issues that occur in analyses in Particle Physics and neighbouring fields. The homepage of PHYSTAT, with a list of all Workshops, Seminars, and Informal Reviews, is at https://phystat.github.io/Website/ .
