Login

Decision on LUT format for PID for upcoming TDR Physics Performance Studies

→ US/Eastern
Description

Zoom info

audio1113435880.m4a
chat.txt
video1113435880.mp4
    • 08:30 → 08:40
      Introduction 10m
    • 08:40 → 09:00
      Contributions / Open mic 20m
      Speaker: TBD
      ayk-delphes-tables-in-athena-and-epic.pdf
      • Log-likelihood based PID interface and fast PID parameterization 5m

        In the proposal era, a (fast) Gaussian response model to translate track-kinematic-dependent PID-sigma-separation provided by RICH groups into the log-likelihood value for hadron PID analysis.

        Here is a presentation about it, containing both its ECCE implementation and the use of log-likelihood PID in LHCb resonance analysis:
        https://indico.bnl.gov/event/13060/#1-hadron-pid

        The code is open-sourced: with a unified PID analysis interface at https://ecce-eic.github.io/doxygen/d2/d02/classEICPIDParticle.html and fast detector parameterization https://github.com/ECCE-EIC/ecce-detectors/tree/master/FastPID for RICH-type detectors. In the proposal time, TOF info was implemented externally to this framework but in principle can be incorporated in a similar fashion.

        One advantage of using log-likelihood PID is to sum information efficiently over multiple PID detectors: When multiple detectors provide PID and their data is independent information, the log-likelihood can be simply summed to provide the global PID Log Likelihood, reflecting the math of independent joint probability distribution from multiple detectors. Here is the code for the sum over the subsystem:
        https://ecce-eic.github.io/doxygen/d0/d8e/EICPIDParticlev1_8cc_source.html#l00066 float EICPIDParticlev1::get_SumLogLikelyhood(EICPIDDefs::PIDCandidate pid) const

        Hope this is helpful

        Speaker: Dr Jin Huang (Brookhaven National Lab)
        02_09_2024_CCPID_EICRECON.pdf
        Past presentation

        In the proposal era, a (fast) Gaussian response model to translate track-kinematic-dependent PID-sigma-separation provided by RICH groups into the log-likelihood value for hadron PID analysis. 

         

        Here is a presentation about it, containing both its ECCE implementation and the use of log-likelihood PID in LHCb resonance analysis: https://indico.bnl.gov/event/13060/#1-hadron-pid

         

        The code is open-sourced: with a unified PID analysis interface at https://ecce-eic.github.io/doxygen/d2/d02/classEICPIDParticle.html and fast detector parameterization https://github.com/ECCE-EIC/ecce-detectors/tree/master/FastPID for RICH-type detectors. In the proposal time, TOF info was implemented externally to this framework but in principle can be incorporated in a similar fashion.

         

        One advantage of using log-likelihood PID is to sum information efficiently over multiple PID detectors: When multiple detectors provide PID and their data is independent information, the log-likelihood can be simply summed to provide the global PID Log Likelihood, reflecting the math of independent joint probability distribution from multiple detectors. Here is the code for the sum over the subsystem: 
        https://ecce-eic.github.io/doxygen/d0/d8e/EICPIDParticlev1_8cc_source.html#l00066  float EICPIDParticlev1::get_SumLogLikelyhood(EICPIDDefs::PIDCandidate pid) const

        Hope this is helpful

    • 09:00 → 09:30
      Discussion 30m
Powered by Indico v3.3.8