We propose a neural network training method capable of accounting for the effects of systematic variations of the data model in the training process and describe its extension towards neural network multiclass classification. The procedure is evaluated on the realistic case of the measurement of Higgs boson production via gluon fusion and vector boson fusion in the ττ decay channel at the CMS experiment. The neural network output functions are used to infer the signal strengths for inclusive production of Higgs bosons as well as for their production via gluon fusion and vector boson fusion. We observe improvements of 12 and 16% in the uncertainty in the signal strengths for gluon and vector-boson fusion, respectively, compared with a conventional neural network training based on cross-entropy.

We propose a neural network training method capable of accounting for the effects of systematic variations of the data model in the training process and describe its extension towards neural network multiclass classification. The procedure is evaluated on the realistic case of the measurement of Higgs boson production via gluon fusion and vector boson fusion in the ττ decay channel at the CMS experiment. The neural network output functions are used to infer the signal strengths for inclusive production of Higgs bosons as well as for their production via gluon fusion and vector boson fusion. We observe improvements of 12 and 16% in the uncertainty in the signal strengths for gluon and vector-boson fusion, respectively, compared with a conventional neural network training based on cross-entropy.

Development of systematic uncertainty-aware neural network trainings for binned-likelihood analyses at the LHC

Alexe, C.;Bruschini, D.;Ligabue, F.;
2025

Abstract

We propose a neural network training method capable of accounting for the effects of systematic variations of the data model in the training process and describe its extension towards neural network multiclass classification. The procedure is evaluated on the realistic case of the measurement of Higgs boson production via gluon fusion and vector boson fusion in the ττ decay channel at the CMS experiment. The neural network output functions are used to infer the signal strengths for inclusive production of Higgs bosons as well as for their production via gluon fusion and vector boson fusion. We observe improvements of 12 and 16% in the uncertainty in the signal strengths for gluon and vector-boson fusion, respectively, compared with a conventional neural network training based on cross-entropy.
2025
Settore PHYS-01/A - Fisica sperimentale delle interazioni fondamentali e applicazioni
   Advanced Multi-Variate Analysis for New Physics Searches at the LHC
   AMVA4NewPhysics
   European Commission
   Horizon 2020 Framework Programme
   675440

   Search for Higgs bosons decaying to charm quarks
   HIGCC
   European Commission
   Horizon 2020 Framework Programme
   724704

   Direct and indirect searches for new physics in events with top quarks using LHC proton-proton collisions at the CMS detector
   LHCTOPVLQ
   European Commission
   Horizon 2020 Framework Programme
   752730

   Majorana neutrino discovery strategy with CMS
   MajorNet
   European Commission
   Horizon 2020 Framework Programme
   758316

   International Training Network for Statistics in High Energy Physics and Society
   INSIGHTS
   European Commission
   Horizon 2020 Framework Programme
   765710

   The strong interaction at the frontier of knowledge: fundamental research and applications
   STRONG-2020
   European Commission
   Horizon 2020 Framework Programme
   824093

   INnovative TRiggEr techniques for beyond the standard model PhysIcs Discovery at the LHC
   INTREPID
   European Commission
   Horizon Europe Framework Programme
   101115353

   Fundamental properties and time-scan of QCD matter at high densities and temperature exposed by jet substructure in heavy ion collisions with CMS experiment at the LHC
   QCDHighDensityCMS
   European Commission
   Horizon 2020 Framework Programme
   101002207
File in questo prodotto:
File Dimensione Formato  
s10052-025-14713-w.pdf

accesso aperto

Tipologia: Published version
Licenza: Creative Commons
Dimensione 2.85 MB
Formato Adobe PDF
2.85 MB Adobe PDF

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11384/159565
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
  • OpenAlex 0
social impact