Daniel Conde Villatoro: Interpretability in machine learning models using symbolic regression: Angular coefficients in Z production

por Daniel Conde Villatoro (Instituto de Física Corpuscular)

martes 28 abr. 2026 15:30 → 17:30 Europe/Madrid

1001-Primera-1-1-1 - Paterna. Seminario (Universe)

Computing angular coefficients for W and Z bosons at the LHC is computationally expensive and often yields results that are unwieldy to use analytically. In this talk, I show how symbolic regression offers a practical alternative: lightweight analytical formulas derived directly from NLO simulation data. Using the PySR package, we recover closed-form expressions for the full set of angular coefficients as functions of transverse momentum, rapidity, and invariant mass. Beyond efficiency, these expressions shed light on the underlying physics in a way that neural networks cannot. Our results demonstrate that symbolic regression can produce accurate and generalisable expressions that match Monte Carlo predictions within uncertainties, while preserving interpretability and providing insight into the kinematic dependence of angular observables.