Apples to Apples in Jet Quenching: robustness of Machine Learning classification of quenched jets to Underlying Event contamination

TL;DR

This paper evaluates the robustness of machine learning classifiers in identifying jet quenching effects in heavy-ion collisions, accounting for underlying event contamination and demonstrating the stability of simple Boosted Decision Trees in realistic scenarios.

Contribution

It establishes a baseline for jet quenching studies that includes underlying event effects and assesses the robustness of ML classifiers against background contamination.

Findings

Boosted Decision Tree classifiers remain effective with underlying event contamination.

Jet quenching observables are robust against background effects in ML classification.

The methodology supports the portability of ML techniques to experimental data.

Abstract

Progress in the theoretical understanding of parton branching dynamics within an expanding Quark Gluon Plasma relies on detailed and fair comparisons with experimental data for reconstructed jets. Such comparisons are only meaningful when the computed jet, be it analytically or via event generation, accounts for the complexity of jets reconstructed in the challenging environment of heavy-ion collisions. Jet reconstruction in heavy ion collisions involves a necessarily imperfect subtraction of the large and fluctuating underlying event: reconstructed jets always include underlying event contamination. To identify true jet quenching effects, modifications due to the interaction of the branching partonic system with the Quark Gluon Plasma, we establish a baseline that accounts for possible background contamination on unmodified jets. In practical terms, jet quenching effects are only those not present in jets produced in proton-proton collisions that have been embedded in a realistic heavy-ion background and where subtraction has been carried out analogously to that in the heavy ion case. With this setup, we assess the sensitivity to underlying event of commonly discussed jet quenching observables and its impact on the robustness of Machine Learning studies, aimed at classifying jets according to their degree of modification by the Quark Gluon Plasma, that rely on those observables. We find the discrimination power of a simple Boosted Decision Tree to be robust in the realistic scenario where both medium response and underlying event are present, giving support to portability to the experimental context.