Probing fundamental interactions at the LHC with machine learning

Tuesday, 30 November 2021 at 11:15

I will introduce the Lund jet plane, a powerful representation of radiation patterns within a jet, and adopt this framework to explore novel machine learning models for jet tagging at the LHC. I will then develop a first-principles understanding of these methods through comparisons with analytic discriminants in specific kinematic limits. I will then discuss limitations of existing dipole showers in accurately describing emissions across multiple energy scales, and show how this might lead to bias in machine learning models. Finally, I will introduce systematic tools to assess and improve the logarithmic accuracy of parton showers.

Slides (PDF, 8 MB)