DeXTer: Deep Set based Neural Networks for boosted multiple b-jet identification in ATLAS

Yuan-Tang Chou^1*, Rafael Coelho Lopes de Sa¹, Rafael Teixeira de Lima², Verena Ingrid Martinez Outschoorn¹, Aurelio Juste Rozas³

¹Department of Physics, University of Massachusetts Amherst, Amherst, MA, USA
²SLAC National Accelerator Laboratory, Stanford, CA, USA
³Institut de Fisica d’Altes Energies (IFAE), The Barcelona Institute of Science and Technology, Barcelona, Spain

* Presenter:Yuan-Tang Chou, email:yuan-tang.chou@cern.ch

This work presents algorithms for flavor tagging identification of jets that are initiated by one or two independent heavy-flavor hadrons. Algorithms in ATLAS for hadronic jets typically focus on high transverse momentum, above 200 GeV. This work describes the first implementation of a double-b tagger for low transverse momentum jets, below 200 GeV. This algorithm relies on large radius track-jets which can be defined at low transverse momenta and implements a deep-set neural network that uses displaced tracks, secondary vertices, and substructure information to identify the presence of multiple heavy-flavored hadrons.

Keywords: B-tagging, Large Hadron Collider, Jet identification