PHYSTAT informal review: GOF tests via Machine Learning

by Alessandra Rosalba Brazzale (University of Padova), Arthur Gretton (Gatsby Computational Neuroscience Unit, UCL), Gaia Grosso (IAIFI, MIT), Dr Marco Letizia (Machine Learning Genoa Center, Univ. of Genoa)

Wednesday 12 Mar 2025, 15:30 → 17:00 Europe/Zurich

This is a PHYSTAT Informal Review event*. Today Gaia Grosso and Marco Letizia  (physicists) together with Alessandra Brazzale and Arthur Gretton (statisticians) will review the topic "Gof tests via ML".

Agenda:

• 3.30 pm Opening:
• 3.30 pm Physicists Presentation (20'+10')
• 4 pm Statisticians Presentation (25'+10')
• 4.35 pm General Discussion and Closing (25')

 

*PHYSTAT informal reviews: In this virtual format, a Tandem consisting of a physicist and a statistician will review a statistical method introduced by one of the parties or a general critical analysis topic from the Physicist's and Statistician's perspectives. The virtual events comprise: two 20+10 min. complementary presentations followed by ~30 minutes of general discussion.

Abstract:

This PHYSTAT-Informal Review focuses on Machine Learning advancements for goodness-of-fit with emphasis on their applications in high-energy physics. Marco Letizia and Gaia Grosso will explore how machine learning enables flexible, non-parametric goodness-of-fit methodologies to overcome challenges posed by Large Hadron Collider datasets, reducing dependence on prior assumptions that can limit anomaly detection. Arthur Gretton will discuss a family of goodness-of-fit tests based on Maximum Mean Discrepancy and popular in the Machine Learning literature. Two adaptive approaches --aggregation and fusing-- that automatically select optimal tests depending on data characteristics will be demonstrated. Alessandra Brazzale will connect these modern approaches to classical statistical foundations, examining how the Neyman-Pearson lemma and likelihood ratio tests relate to the New Physics Learning Machine.

 

 

More on "Classical GOF tests":

Bob Cousins' lectures on statistics in Theory:

https://arxiv.org/abs/1807.05996

Wolfgang Rolke's PHYSTAT seminar on GOF tests:

https://indico.cern.ch/event/972293/

Alessandra_Brazzale.pdf