Performance versus resilience in modern quark-gluon tagging

Butter, Anja; Dillon, Barry M.; Plehn, Tilman; Vogel, Lorenz

doi:10.21468/SciPostPhysCore.6.4.085

SciPost Physics Core

Performance versus resilience in modern quark-gluon tagging

Anja Butter, Barry M. Dillon, Tilman Plehn, Lorenz Vogel

SciPost Phys. Core 6, 085 (2023) · published 12 December 2023

doi: 10.21468/SciPostPhysCore.6.4.085
pdf
Submissions/Reports

Abstract

Discriminating quark-like from gluon-like jets is, in many ways, a key challenge for many LHC analyses. First, we use a known difference in PYTHIA and HERWIG simulations to show how decorrelated taggers would break down when the most distinctive feature is aligned with theory uncertainties. We propose conditional training on interpolated samples, combined with a controlled Bayesian network, as a more resilient framework. The interpolation parameter can be used to optimize the training evaluated on a calibration dataset, and to test the stability of this optimization. The interpolated training might also be useful to track generalization errors when training networks on simulations.

TY  - JOUR
PB  - SciPost Foundation
DO  - 10.21468/SciPostPhysCore.6.4.085
TI  - Performance versus resilience in modern quark-gluon tagging
PY  - 2023/12/12
UR  - https://scipost.org/SciPostPhysCore.6.4.085
JF  - SciPost Physics Core
JA  - SciPost Phys. Core
VL  - 6
IS  - 4
SP  - 085
A1  - Butter, Anja
AU  - Dillon, Barry M.
AU  - Plehn, Tilman
AU  - Vogel, Lorenz
AB  - Discriminating quark-like from gluon-like jets is, in many ways, a key challenge for many LHC analyses. First, we use a known difference in PYTHIA and HERWIG simulations to show how decorrelated taggers would break down when the most distinctive feature is aligned with theory uncertainties. We propose conditional training on interpolated samples, combined with a controlled Bayesian network, as a more resilient framework. The interpolation parameter can be used to optimize the training evaluated on a calibration dataset, and to test the stability of this optimization. The interpolated training might also be useful to track generalization errors when training networks on simulations.
ER  -

@Article{10.21468/SciPostPhysCore.6.4.085,
	title={{Performance versus resilience in modern quark-gluon tagging}},
	author={Anja Butter and Barry M. Dillon and Tilman Plehn and Lorenz Vogel},
	journal={SciPost Phys. Core},
	volume={6},
	pages={085},
	year={2023},
	publisher={SciPost},
	doi={10.21468/SciPostPhysCore.6.4.085},
	url={https://scipost.org/10.21468/SciPostPhysCore.6.4.085},
}

Cited by 6

Authors / Affiliations: mappings to Contributors and Organizations

See all Organizations.

Funders for the research work leading to this publication