lista

Document Sample
lista Powered By Docstoc
					Muon Identification


         L.Lista
  INFN Sezione di Napoli


         Luca Lista
Outline
• … I’ll assume you have seen my plenary talk
• I will focus on aspects where we need feedback from
  PID group about:


• Muon selection with the upgraded forward
• Likelihood selector
    More on Francesco’s talk
• Neural Network selector




                        Luca Lista
New muon selector
• We have to finalize selection:
   Essentially complete for the forward part (Tina)
     • Finalized the cuts based on ntuple studies
   Update the C++ selector code with new cuts
    (Francesco)
   The Belt still shows problems
     • Bugs in the latest production, hopefully fixed
     • New productions going on (Gianluca) .
     • Presumably the selection in the belt will change
       accordingly




                         Luca Lista
Belt problems
p > 1.8 GeV/c          p > 1.8 GeV/c           p > 1.8 GeV/c




                Belt                    Belt                   Belt




• Recently discovered missing Front End Cards
  in the simulation in the conditions db
     Effectively missing part of the detector!
     Should be fixed in new productions

                           Luca Lista
Upgrade selector code
• The appropriate selector is chosen at run time
  according to a key fetched from the conditions
  database
    Will work eventually for the new barrel too!
• We have almost completed the multi-selector code
    Kanga and Objy persistency done (L.)
    Just missing “link” with RooCond (Eugenio)
      • should be straightforward
    Several packages involved:
      • BetaBuildPidEnv, BetaPid, BtaEnv, IfrPidP,
        IfrPidR, IfrPidUtils, IfrProxy
    Don’t expect problems, but need to be coordinated



                              Luca Lista
Composite selector design
                              PidChargedSelector




                       0..*
                               MuonAbsSelector



                                                               MuonNewBarrelSelector
             1
       MuonMicroSelector

                                                    MuonFWDSelector

                                                                                Future Selectors

     IfrDetConfigKey          MuonClassicSelector


                                                                 New Selector

                                         Old Selector
        From Cond. DB




                                        Luca Lista
Likelihood selector
• Code ready and usable (Francesco)
• Improved performances
• Need to reload the constants in the database for some
  of the periods
    Must be done before we produce PID tables
• Still more tuning, or better to release as it is?
• PID tables should be produced with the selector
    Latest tag of BetaPidCalib stores likelihoods and new
     selector bit in order to produce performance tables
    Need special care: sub-sample not used for PDF “training”
      • even run number only



                            Luca Lista
Likelihood selector
                             Further
                             tuning?




                Luca Lista
Neural Networks
• Good improvements are expected (Ajit)
• It would be possible to provide different selectors for
  different analyses needs
• Tune according to:
    Constant Muon efficiency
    Constant Pion misid
    Constant Muon / Pion ratio
• The selector code and the “infrastructure” have still to
  be implemented
• We should decide which priority should we give to:
    release of the existing
    selection improvements


                               Luca Lista
Conclusions
• Finalizing the new forward selector
• Releasing Likelihood selector
• Trying to boost Neural Networks

• … PID team, please provide feedback!




                     Luca Lista

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:10
posted:8/12/2012
language:
pages:10