Learning to Isolate Muons in Data

TL;DR

This paper employs weakly-supervised learning on LHC collision data to improve muon isolation techniques by utilizing low-level particle activity patterns, surpassing traditional scalar isolation measures.

Contribution

It introduces a data-driven approach using Classification Without Labels to identify prompt muons, revealing that energy flow polynomials can enhance discrimination beyond standard isolation.

Findings

Isolation is less effective than low-level calorimeter information.

Energy flow polynomials can close the performance gap in muon identification.

Training on real data yields different polynomials than those from simulation.

Abstract

We use unlabeled collision data and weakly-supervised learning to train models which can distinguish prompt muons from non-prompt muons using patterns of low-level particle activity in the vicinity of the muon, and interpret the models in the space of energy flow polynomials. Particle activity associated with muons is a valuable tool for identifying prompt muons, those due to heavy boson decay, from muons produced in the decay of heavy flavor jets. The high-dimensional information is typically reduced to a single scalar quantity, isolation, but previous work in simulated samples suggests that valuable discriminating information is lost in this reduction. We extend these studies in LHC collisions recorded by the CMS experiment, where true class labels are not available, requiring the use of the invariant mass spectrum to obtain macroscopic sample information. This allows us to employ Classification Without Labels (CWoLa), a weakly supervised learning technique, to train models. Our results confirm that isolation does not describe events as well as the full low-level calorimeter information, and we are able to identify single energy flow polynomials capable of closing the performance gap. These polynomials are not the same ones derived from simulation, highlighting the importance of training directly on data.