Exploring the Universality of Hadronic Jet Classification

TL;DR

This study finds that machine learning classifiers for jet substructure are nearly universal across different simulation models, suggesting robust performance when trained on one and tested on another or real data.

Contribution

It demonstrates the near-universality of ML classifiers across different PSMC models in jet classification, highlighting their robustness and transferability.

Findings

Classifiers trained on one PSMC perform similarly on others.

Neural networks show consistent performance across simulations.

Results suggest potential for reliable application to real data.

Abstract

The modeling of jet substructure significantly differs between Parton Shower Monte Carlo (PSMC) programs. Despite this, we observe that machine learning classifiers trained on different PSMCs learn nearly the same function. This means that when these classifiers are applied to the same PSMC for testing, they result in nearly the same performance. This classifier universality indicates that a machine learning model trained on one simulation and tested on another simulation (or data) will likely be optimal. Our observations are based on detailed studies of shallow and deep neural networks applied to simulated Lorentz boosted Higgs jet tagging at the LHC.