BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:UW-Madison-Physics-Events
BEGIN:VEVENT
SEQUENCE:2
UID:UW-Physics-Event-8746
DTSTART:20240510T140000Z
DTEND:20240510T160000Z
DTSTAMP:20260413T205113Z
LAST-MODIFIED:20240503T144439Z
LOCATION:https://uwmadison.zoom.us/j/94127401731?pwd=V1drQ29XeFJMTkRzV
 Ul0OExobVJzZz09
SUMMARY:Application of geometric deeping learning in charged-particle 
 track reconstruction in the ATLAS ITk\, Preliminary Exam\, Tuan Pham\,
  Physics Graduate Student
DESCRIPTION:The reconstruction of charged-particle trajectories\, rank
 ing amongst the most computationally demanding tasks in particle colli
 der experiments\, such as the ATLAS experiment at CERN\, plays an esse
 ntial role in any High-Energy Physics program\, as it determines the q
 uality of particle identification\, kinematic measurement\, vertex fin
 ding\, lepton reconstruction\, jet flavor tagging\, and other downstre
 am tasks. The upcoming High Luminosity phase of the Large Hadron Colli
 der (HL-LHC) represents a steep increase in the average number of prot
 on-proton interactions and hence in the computing resources required f
 or offline track reconstruction of the ATLAS Inner Tracker (ITk). As s
 uch\, track pattern recognition algorithms based on Graph Neural Netwo
 rks (GNNs) have been demonstrated as a promising approach to these cha
 llenges. We present a novel algorithm developed for track reconstructi
 on in silicon detectors based on a number of deep learning techniques 
 including GNN architectures. Using detector simulation of collision ev
 ents associated with the production of a top quark pair on the latest 
 version of ITk geometry under HL-LHC conditions\, we demonstrate the p
 erformance of our algorithm\, and compare to that of the tracking algo
 rithm currently used in ATLAS on a range of important physics metrics\
 , including reconstruction efficiency\, and track parameter resolution
 . Finally\, we discuss the algorithm's computational performance and o
 ptimisations that reduce computing costs\, as well as our effort to in
 tegrate into the ATLAS analysis software for full-chain testing and pr
 oduction.
URL:https://www.physics.wisc.edu/events/?id=8746
END:VEVENT
END:VCALENDAR