Studies to mitigate difference between real data and simulation for jet tagging

Mostrar otras versiones (4)

<!--HTML-->The aim of the studies presented is to improve the performance of jet flavour tagging on real data while still exploiting a simulated dataset for the learning of the main classification task. In the presentation we explore “off the shelf” domain adaptation techniques as well as cust...

Descripción completa

Detalles Bibliográficos
Autores principales: Buchmuller, Oliver, Martelli, Arabella, Kieseler, Jan, Verzetti, Mauro, Stoye, Markus
Lenguaje:eng
Publicado: 2018
Materias:
Machine Learning
Acceso en línea:http://cds.cern.ch/record/2315380
_version_ 1780958141094035456
author Buchmuller, Oliver
Martelli, Arabella
Kieseler, Jan
Verzetti, Mauro
Stoye, Markus
author_facet Buchmuller, Oliver
Martelli, Arabella
Kieseler, Jan
Verzetti, Mauro
Stoye, Markus
author_sort Buchmuller, Oliver
collection CERN
description <!--HTML-->The aim of the studies presented is to improve the performance of jet flavour tagging on real data while still exploiting a simulated dataset for the learning of the main classification task. In the presentation we explore “off the shelf” domain adaptation techniques as well as customised additions to them. The latter improves the calibration of the tagger, potentially leading to smaller systematic uncertainties. The studies are performed with simplified simulations for the case of b-jet tagging. The presentation will include first results as well as discuss pitfalls that we discovered during our research.
id cern-2315380
institution Organización Europea para la Investigación Nuclear
language eng
publishDate 2018
record_format invenio
spelling cern-23153802022-11-02T22:34:02Zhttp://cds.cern.ch/record/2315380engBuchmuller, OliverMartelli, ArabellaKieseler, JanVerzetti, MauroStoye, MarkusStudies to mitigate difference between real data and simulation for jet tagging2nd IML Machine Learning WorkshopMachine Learning<!--HTML-->The aim of the studies presented is to improve the performance of jet flavour tagging on real data while still exploiting a simulated dataset for the learning of the main classification task. In the presentation we explore “off the shelf” domain adaptation techniques as well as customised additions to them. The latter improves the calibration of the tagger, potentially leading to smaller systematic uncertainties. The studies are performed with simplified simulations for the case of b-jet tagging. The presentation will include first results as well as discuss pitfalls that we discovered during our research.oai:cds.cern.ch:23153802018
spellingShingle Machine Learning
Buchmuller, Oliver
Martelli, Arabella
Kieseler, Jan
Verzetti, Mauro
Stoye, Markus
Studies to mitigate difference between real data and simulation for jet tagging
title Studies to mitigate difference between real data and simulation for jet tagging
title_full Studies to mitigate difference between real data and simulation for jet tagging
title_fullStr Studies to mitigate difference between real data and simulation for jet tagging
title_full_unstemmed Studies to mitigate difference between real data and simulation for jet tagging
title_short Studies to mitigate difference between real data and simulation for jet tagging
title_sort studies to mitigate difference between real data and simulation for jet tagging
topic Machine Learning
url http://cds.cern.ch/record/2315380
work_keys_str_mv AT buchmulleroliver studiestomitigatedifferencebetweenrealdataandsimulationforjettagging
AT martelliarabella studiestomitigatedifferencebetweenrealdataandsimulationforjettagging
AT kieselerjan studiestomitigatedifferencebetweenrealdataandsimulationforjettagging
AT verzettimauro studiestomitigatedifferencebetweenrealdataandsimulationforjettagging
AT stoyemarkus studiestomitigatedifferencebetweenrealdataandsimulationforjettagging
AT buchmulleroliver 2ndimlmachinelearningworkshop
AT martelliarabella 2ndimlmachinelearningworkshop
AT kieselerjan 2ndimlmachinelearningworkshop
AT verzettimauro 2ndimlmachinelearningworkshop
AT stoyemarkus 2ndimlmachinelearningworkshop

Ejemplares similares