Electron Identification using Likelihood

Complete: 4

Introduction

The following variables are currently used to discriminate between real and fake electrons in the likelihood:

  • Energy of closest BasicCluster to track impact point at ECAL / outermost track momentum
  • DeltaEta between SuperCluster position and track direction at vertex extrapolated to ECAL assuming no radiation
  • DeltaPhi between SuperCluster position and track direction at vertex
  • Ratio of energy in HCAL behind SuperCluster to SuperCluster energy
  • Combination in a linear Fisher discriminant of cluster shape variables:
    • Energy in 3x3 crystals / energy in 5x5 crystals
    • SigmaEtaEta cluster shape covariance
    • Lateral moment
    • Complex Zernike moment A20

Specialized pdfs are for:

  • ECAL barrel, ECAL endcap
  • electron pT < 15GeV / >15GeV
  • electron classes (only for signal). Two possibilities are allowed:
    • split by "fullclass": golden, bigbrem, narrow, showering
    • split by "class": non-showering (golden+bigbrem+narrow all together weighted by their fractions), showering

Species: only two species are currently considered:

  • Electrons: pdfs taken from pure di-electron sample, particle gun
  • Jets: QCD samples, pdfs are the weighted sum of pdf's for jet pT bins from 0-230 GeV. The weight is the product of the reconstruction efficiency as a pixelMatchGsfElectron x cross section.

Configuration

Configuration can be done changing the .cfi file:

  • EgammaAnalysis/ElectronIDESSources/data/likelihoodESetup.cfi

In the default configuration, signal species (electrons) pdf's are split by "fullclass" and background species is not split. For the signal species, the possibilities are:

  • split by "fullclass". The configuration is:
    string signalWeightSplitting = "fullclass"
    bool splitSignalPdfs = true
  • split by "class":
    string signalWeightSplitting = "class"
    bool splitSignalPdfs = true
  • do not split at all:
    string signalWeightSplitting = "fullclass"
    bool splitSignalPdfs = false

In principle also for background the same procedure is available. Due to the low MC statistics available in the Bigbrem/Narrow classes, right now only the unsplit pdf's are available in the DB. Configuration is:

  • do not split at all:
    string backgroundWeightSplitting = "fullclass"
    bool splitBackgroundPdfs = false

Setup of the code

compilation in 17X

cvs co -r V00-00-07      EgammaAnalysis/ElectronIDAlgos
cvs co -r V00-00-04      EgammaAnalysis/ElectronIDProducers
cvs co -r V00-00-03      EgammaAnalysis/ElectronIDESSources
cvs co -r V00-01-01      CondCore/EgammaPlugins                           
cvs co -r V00-02-00      CondFormats/EgammaObjects 

compilation in 16X

cvs co -r V00-00-07      EgammaAnalysis/ElectronIDAlgos
cvs co -r V00-00-04      EgammaAnalysis/ElectronIDProducers
cvs co -r V00-00-03      EgammaAnalysis/ElectronIDESSources
cvs co -r V00-01-01      CondCore/EgammaPlugins
cvs co -r V02-06-02      CondFormats/DataRecord
cvs co -r V00-02-00      CondFormats/EgammaObjects
cvs co -r V00-01-03      CondFormats/PhysicsToolsObjects
cvs co -r V00-05-00      DataFormats/EgammaReco
cvs co -r V00-00-03      PhysicsTools/StatPatternRecognition
cvs co -r V00-05-07      RecoEcal/EgammaClusterProducers
cvs co -r V00-03-00      RecoEcal/EgammaCoreTools

Since the SQLite DB file is only available in releases >17X, use a local file. Change, in EgammaAnalysis/ElectronIDESSources/data/likelihoodPdfsDB.cfi

replace CondDBCommon.connect = "sqlite_file:/afs/cern.ch/user/e/emanuele/scratch0/vanilla131HLT6/src/EgammaAnalysis/ElectronIDAlgos/test/electronlikelihood.db"    
replace CondDBCommon.catalog = "file:/afs/cern.ch/user/e/emanuele/scratch0/vanilla131HLT6/src/EgammaAnalysis/ElectronIDAlgos/test/mycatalog.xml"

with:

replace CondDBCommon.connect = "sqlite_file:/afs/cern.ch/user/e/emanuele/public/electronLikelihoodDb160/electronIdLikelihood.db"    
replace CondDBCommon.catalog = "file:/afs/cern.ch/user/e/emanuele/public/electronLikelihoodDb160/likelihoodcatalog.xml"

compilation in 131HLT6

cvs co -r V00-00-04      EgammaAnalysis/ElectronIDAlgos
cvs co -r V00-00-01      EgammaAnalysis/ElectronIDESSources
cvs co -r V00-05-00      DataFormats/EgammaReco
cvs co -r V00-01-00      CondCore/EgammaPlugins
cvs co -r V00-01-10      CondFormats/BTauObjects
cvs co -r V02-06-02      CondFormats/DataRecord
cvs co -r V00-01-00      CondFormats/EgammaObjects
cvs co -r V00-03-04      RecoEcal/EgammaClusterAlgos
cvs co -r V00-05-07      RecoEcal/EgammaClusterProducers
cvs co -r V00-03-00      RecoEcal/EgammaCoreTools
cvs co -r shr-8aug07     PhysicsTools/StatPatternRecognition

Write your own Analyzer or Producer

Examples of analyzers tested in CMSSW_160 can be found here:

  • Analyzer: /afs/cern.ch/user/e/emanuele/public/4Likelihood/LikelihoodAnalyzer160.tar.gz
  • Producer: /afs/cern.ch/user/e/emanuele/public/4Likelihood/ElectronIDProducers160.tar.gz
the last one produces a collection of references to only the electrons which satisfy cut on the likelihood output configurable by the .cfg.

-- EmanueleDiMarco - 08 Oct 2007

Edit | Attach | Watch | Print version | History: r16 | r9 < r8 < r7 < r6 | Backlinks | Raw View | Raw edit | More topic actions...
Topic revision: r7 - 2008-02-07 - RobertoSalerno
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    CMSPublic All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback