def MakeConf(Verbose=True):
parser = argparse.ArgumentParser()
parser.add_argument(
'--inputrootfile',
'-i',
type=str,
help='Path to root file',
# default='../applyRegression/Config_Sep30_electron_EB_ECALonly_appliedRegression_training.root'
# default='../Config_Oct25_electron_EB_ECALonly_appliedRegression_training_ptWeight.root'
)
parser.add_argument('--region',
metavar='N',
type=str,
nargs='+',
help='Specify regions',
default=['EB', 'EE'],
choices=['EE', 'EB'])
parser.add_argument(
'-n',
'--name',
type=str,
default='NONE',
help='Append a string at the end of the name of this config')
parser.add_argument(
'--fast',
action='store_true',
help='Change some BDT options to be faster (but maybe less precise)')
args = parser.parse_args()
datestr = strftime('%b%d')
# Photon does not have TRK vars
particle = 'electron'
# Reads off the name of this .py file, so it's clear what made this.
moduleName = os.path.basename(__file__).replace('MakeConf_',
'').replace('.py', '')
return_configs = []
for region in args.region:
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = 'Config_' + datestr + '_' + particle + '_' + region + '_' + moduleName
# Append a string to the name if given by the user
if not args.name == 'NONE':
config.Name += '_' + args.name
if args.fast: config.Name += '_FastOptions'
config.InputFiles = os.path.abspath(args.inputrootfile)
# config.Tree = 'een_analyzer/{0}Tree'.format( particle.capitalize() )
config.Tree = 'een_analyzer/correction' # <-- May want to change this some time.
########################################
# BDT settings
########################################
if not args.fast:
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=1",
]
else:
config.Options = [
"MinEvents=300", # Down from 200
"Shrinkage=0.2",
"NTrees=1000",
"MinSignificance=5.0", # Down from 5.0
"EventWeight=1",
]
# config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.Target = "(genEnergy * (trkMomentum*trkMomentum*trkMomentumRelError*trkMomentumRelError + (scRawEnergy+scPreshowerEnergy)*(scRawEnergy+scPreshowerEnergy)*resolution*resolution) / ( (scRawEnergy+scPreshowerEnergy)*response*trkMomentum*trkMomentum*trkMomentumRelError*trkMomentumRelError + trkMomentum*(scRawEnergy+scPreshowerEnergy)*(scRawEnergy+scPreshowerEnergy)*resolution*resolution ))"
# Probably neither of these are necessary
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
if region == 'EB':
config.DoEB = "True"
else:
config.DoEB = "False"
# # ======================================
# # Sample division - need a part for the ECAL-only training, and a part for the combination
# # 80% for the main BDT - divide the sample in divideNumber pieces, and use all but one piece for the main BDT
# divideNumber = 3
# config.CutBase = "eventNumber%{0}!=0".format( divideNumber )
# # 10% for combination, 10% for error
# config.CutComb = "eventNumber%{0}==0 && eventNumber%{1}==0".format( divideNumber, 2*divideNumber )
# config.CutError = "eventNumber%{0}==0 && eventNumber%{1}!=0".format( divideNumber, 2*divideNumber )
# config.CutBase += " && NtupID<5000"
# config.CutComb += " && NtupID<5000"
# config.CutError += " && NtupID<5000"
# Limit number of events in training
# config.CutBase = "NtupID<1000"
# config.CutComb = "NtupID<1000"
# config.CutError = "NtupID<1000"
# Pre-selected events have this variable set to 1.0
config.CutBase = "(ptWeightCut)"
config.CutComb = "(ptWeightCut)"
config.CutError = "(ptWeightCut)"
########################################
# Order tree branches
########################################
common_vars = [
"(scRawEnergy+scPreshowerEnergy)*response",
"resolution/response",
"trkMomentumRelError",
"trkMomentum/((scRawEnergy+scPreshowerEnergy)*response)",
"eleEcalDriven",
"fbrem",
"gsfchi2",
"gsfndof",
"trkEta",
"trkPhi",
]
config.VariablesEB = common_vars + []
config.VariablesEE = common_vars + []
if Verbose:
print '\n' + '-' * 70
print 'Making config file ' + config.Name + '.config'
print ' Using the following branches for EE:'
print ' ' + '\n '.join(config.VariablesEE)
print ' Using the following branches for EB:'
print ' ' + '\n '.join(config.VariablesEB)
config.DoCombine = "False"
config.DoErrors = "False"
########################################
# Output
########################################
config.Parse()
return_configs.append(config)
return return_configs
def main():
# Small testing samples -- do NOT use these for plots!
# fullpt_root_file = 'Ntup_Jun22_fullpt_testing_sample.root'
# lowpt_root_file = 'Ntup_Jun22_lowpt_testing_sample.root'
# Low + high pt sample
fullpt_root_file = 'Ntup_Jun22_fullpt_training.root'
# Only low pt sample
lowpt_root_file = 'Ntup_Jun22_lowpt_training.root'
ntup_path = os.path.join( os.environ['CMSSW_BASE'], 'src/NTuples' )
datestr = strftime( '%b%d' )
if not os.path.isdir( ntup_path ):
print 'Error: "{0}"" is not a directory'.format( ntup_path )
physical_path = lambda input_root_file: os.path.join( ntup_path, input_root_file )
########################################
# BASE CONFIG - This is low pt electrons
# Configs for photons and and other pt ranges are created by altering this one
########################################
# Instantiate the Config class which prints a .config file
base_config = Config()
base_config.Name = 'Config_electron_lowpt_' + datestr
base_config.InputFiles = physical_path( lowpt_root_file )
base_config.Tree = 'een_analyzer/ElectronTree'
########################################
# BDT settings
########################################
base_config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=max( min(1,exp(-(genPt-50)/50)), 0.1 )", # <-- What to do?
]
base_config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
base_config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
base_config.HistoConfig = "jobs/dummy_Histo.config"
base_config.CutBase = "eventNumber%2==0"
base_config.CutEB = "scIsEB"
base_config.CutEE = "!scIsEB"
base_config.CutError = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4==3)"
# Add an additional cut so that the regression is fast
# NtupIDcut = 10000
# base_config.CutBase += ' && (NtupID<={0})'.format( NtupIDcut )
# base_config.CutError += ' && (NtupID<={0})'.format( NtupIDcut )
# base_config.CutComb += ' && (NtupID<={0})'.format( NtupIDcut )
########################################
# Order tree branches
########################################
common_vars = [
# ======================================
# Common variables
'pt',
# 'nVtx', # rho should be enough information for the BDT
'scRawEnergy',
# 'scEta', # Requires alignment information; use crystal number of the seed instead
# 'scPhi', # Requires alignment information; use crystal number of the seed instead
'scEtaWidth',
'scPhiWidth',
'scSeedRawEnergy/scRawEnergy',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
# ======================================
# Showershape variables
# Use full 5x5 instead
# 'r9',
# 'eHorizontal',
# 'eVertical',
# 'sigmaIetaIeta',
# 'sigmaIetaIphi',
# 'sigmaIphiIphi',
# 'e5x5',
# 'e3x3',
# 'eMax',
# 'e2nd',
# 'eTop',
# 'eBottom',
# 'eLeft',
# 'eRight',
# 'e2x5Max',
# 'e2x5Left',
# 'e2x5Right',
# 'e2x5Top',
# 'e2x5Bottom',
# Normalization to scRawEnergy necessary?
'full5x5_r9',
'full5x5_eHorizontal',
'full5x5_eVertical',
'full5x5_sigmaIetaIeta',
'full5x5_sigmaIetaIphi',
'full5x5_sigmaIphiIphi',
'full5x5_e5x5',
'full5x5_e3x3',
'full5x5_eMax',
'full5x5_e2nd',
'full5x5_eTop',
'full5x5_eBottom',
'full5x5_eLeft',
'full5x5_eRight',
'full5x5_e2x5Max',
'full5x5_e2x5Left',
'full5x5_e2x5Right',
'full5x5_e2x5Top',
'full5x5_e2x5Bottom',
# ======================================
# Saturation variables
'N_SATURATEDXTALS',
'seedIsSaturated',
'seedCrystalEnergy/scRawEnergy',
# ======================================
# Cluster variables
'N_ECALClusters',
'clusterMaxDR',
'clusterMaxDRDPhi',
'clusterMaxDRDEta',
'clusterMaxDRRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
base_config.VariablesEB = common_vars + [
# 'cryEtaCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryPhiCoordinate', # Requires alignment information; use crystal number of the seed instead
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
base_config.VariablesEE = common_vars + [
# 'cryXCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryYCoordinate', # Requires alignment information; use crystal number of the seed instead
'iXCoordinate',
'iYCoordinate',
'scPreshowerEnergy/scRawEnergy',
'preshowerEnergyPlane1/scRawEnergy',
'preshowerEnergyPlane2/scRawEnergy',
]
# print 'Using the following branches for EE:'
# print ' ' + '\n '.join( base_config.VariablesEE )
# print 'Using the following branches for EB:'
# print ' ' + '\n '.join( base_config.VariablesEB )
########################################
# Ep combination
########################################
# Only do the combination for the electron
base_config.DoCombine = "True"
base_config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
base_config.CutComb = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4!=3)"
base_config.VariablesComb = [
'( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
'BDTerror/BDTresponse',
'trkMomentum',
'trkMomentumRelError',
'BDTerror/BDTresponse/trkMomentumRelError',
'( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum',
( '( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum *' +
'sqrt( BDTerror/BDTresponse*BDTerror/BDTresponse + trkMomentumRelError*trkMomentumRelError)' ),
'eleEcalDriven',
'eleTrackerDriven',
'eleClass',
'scIsEB',
]
########################################
# Output
########################################
# lowpt electrons - this is simply the base config defined above
base_config.Parse()
# fullpt electrons - only change the root file
base_config.Name = 'Config_electron_fullpt_' + datestr
base_config.InputFiles = physical_path( fullpt_root_file )
base_config.Parse()
# lowpt photons
base_config.Name = 'Config_photon_lowpt_' + datestr
base_config.InputFiles = physical_path( lowpt_root_file )
base_config.Tree = 'een_analyzer/PhotonTree'
base_config.DoCombine = "False"
base_config.Parse()
# fullpt photons
base_config.Name = 'Config_photon_fullpt_' + datestr
base_config.InputFiles = physical_path( fullpt_root_file )
base_config.Tree = 'een_analyzer/PhotonTree'
base_config.DoCombine = "False"
base_config.Parse()
########################################
# OLD VARIABLES
########################################
# Remove the max( ..., 0.1, ) from the eventweight
base_config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=min(1,exp(-(genPt-50)/50))",
]
# lowpt electrons
base_config.Name = 'Config_electron_lowpt_' + datestr + '_OLDVARS'
base_config.InputFiles = physical_path( lowpt_root_file )
base_config.Tree = 'een_analyzer/ElectronTree'
base_config.DoCombine = "True"
OLD_common_electron_vars = [
'nVtx',
'scRawEnergy',
'scEta',
'scPhi',
'scEtaWidth',
'scPhiWidth',
'r9',
'scSeedRawEnergy/scRawEnergy',
'eMax',
'e2nd',
'eHorizontal', # 'scSeedLeftRightAsym',
'eVertical', # 'scSeedTopBottomAsym',
'sigmaIetaIeta',
'sigmaIetaIphi',
'sigmaIphiIphi',
'N_ECALClusters',
'clusterMaxDR',
'clusterMaxDRDPhi',
'clusterMaxDRDEta',
'clusterMaxDRRawEnergy/scRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
base_config.VariablesEB = OLD_common_electron_vars + [
'cryEtaCoordinate',
'cryPhiCoordinate',
'iEtaCoordinate',
'iPhiCoordinate',
# 'scSeedCryEta',
# 'scSeedCryPhi',
# 'scSeedCryIetaV2',
# 'scSeedCryIphiV2',
]
base_config.VariablesEE = OLD_common_electron_vars + [
'scPreshowerEnergy/scRawEnergy',
# 'scSeedCryIxV2',
# 'scSeedCryIyV2',
'iXCoordinate',
'iYCoordinate',
]
base_config.Parse()
# fullpt oldvars
base_config.Name = 'Config_electron_fullpt_' + datestr + '_OLDVARS'
base_config.InputFiles = physical_path( fullpt_root_file )
base_config.Parse()
# lowpt photons
base_config.Name = 'Config_photon_lowpt_' + datestr + '_OLDVARS'
base_config.InputFiles = physical_path( lowpt_root_file )
base_config.Tree = 'een_analyzer/PhotonTree'
base_config.DoCombine = "False"
OLD_common_photon_vars = [
'nVtx',
'scRawEnergy',
# 'scEta',
# 'scPhi',
'scEtaWidth',
'scPhiWidth',
'r9',
'scSeedRawEnergy/scRawEnergy',
# 'scSeedLeftRightAsym',
# 'scSeedTopBottomAsym',
'sigmaIetaIeta',
'sigmaIetaIphi',
'sigmaIphiIphi',
'N_ECALClusters',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
'e3x3/e5x5',
'eMax/e5x5',
'e2nd/e5x5',
'eTop/e5x5',
'eBottom/e5x5',
'eLeft/e5x5',
'eRight/e5x5',
'e2x5Max/e5x5',
'e2x5Left/e5x5',
'e2x5Right/e5x5',
'e2x5Top/e5x5',
'e2x5Bottom/e5x5',
]
base_config.VariablesEB = OLD_common_photon_vars + [
'e5x5/scSeedRawEnergy',
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
base_config.VariablesEE = OLD_common_photon_vars + [
'scPreshowerEnergy/scRawEnergy',
'preshowerEnergyPlane1/scRawEnergy',
'preshowerEnergyPlane2/scRawEnergy',
'iXCoordinate',
'iYCoordinate',
]
base_config.Parse()
# fullpt photons
base_config.Name = 'Config_photon_fullpt_' + datestr + '_OLDVARS'
base_config.InputFiles = physical_path( fullpt_root_file )
base_config.Parse()
# Print all branches as a check
print "\nAll branches in lowpt root file:"
Read_branches_from_rootfile( physical_path(lowpt_root_file) , base_config.Tree )
def main():
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = "photonConfig"
# filename of the input root file
#root_file = 'FlatNtupFull_13May_SepTrees.root'
#root_file = 'FlatNtupFull_18May_DoubleElectron.root'
#root_file = 'Ntup_20May_DoubleElectron.root'
#root_file = 'Ntup_30May_DoublePhoton_somefailed.root'
#root_file = 'Ntup_01June_DoublePhoton.root'
#root_file = 'Ntup_05June_photons_LowHighPt.root'
root_file = 'Ntup_12June_photons_lowhighpt.root'
ntup_path = os.path.abspath('../../NTuples/')
# For iterating:
#root_file = 'output.root'
#ntup_path = '/afs/cern.ch/work/t/tklijnsm/EGM/CMSSW_8_0_4/src/SimpleFlatTreeProducer/SimpleNtuplizer/cfgs/'
if not os.path.isdir(ntup_path):
print 'Error: "{0}"" is not a directory'.format(ntup_path)
physical_path = lambda input_root_file: os.path.join(
ntup_path, input_root_file)
########################################
# BDT settings
########################################
config.InputFiles = physical_path(root_file)
# Don't run the Ep combination for photons
config.DoCombine = "False"
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=min(1,exp(-(genPt-50)/50))",
]
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
#config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutBase = "eventNumber%2==0 && genPt<2000"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
config.CutError = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4==3)"
#config.CutComb = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4!=3)"
# # Add an additional cut so that the regression is fast
# NtupIDcut = 200
# config.CutBase += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutError += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutComb += ' && (NtupID<={0})'.format( NtupIDcut )
########################################
# Order Electron tree branches
########################################
# Try to read tree branches from the input root file
#tree_gDirectory = 'een_analyzer/ElectronTree'
tree_gDirectory = 'een_analyzer/PhotonTree'
# ---------------------
# JBs branches for EE (order is important)
# ---------------------
common_vars = [
# Same as in electron case
'nVtx',
'scRawEnergy',
# 'scEta',
# 'scPhi',
'scEtaWidth',
'scPhiWidth',
'scSeedR9',
'scSeedRawEnergy/scRawEnergy',
# 'scSeedLeftRightAsym',
# 'scSeedTopBottomAsym',
'scSeedSigmaIetaIeta',
'scSeedSigmaIetaIphi',
'scSeedSigmaIphiIphi',
'N_ECALClusters',
# 'clusterMaxDR',
# 'clusterMaxDRDPhi',
# 'clusterMaxDRDEta',
# 'clusterMaxDRRawEnergy/scRawEnergy',
# 'clusterRawEnergy[0]/scRawEnergy',
# 'clusterRawEnergy[1]/scRawEnergy',
# 'clusterRawEnergy[2]/scRawEnergy',
# 'clusterDPhiToSeed[0]',
# 'clusterDPhiToSeed[1]',
# 'clusterDPhiToSeed[2]',
# 'clusterDEtaToSeed[0]',
# 'clusterDEtaToSeed[1]',
# 'clusterDEtaToSeed[2]',
# Only for photons
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
'e3x3/e5x5',
'eMax/e5x5',
'e2nd/e5x5',
'eTop/e5x5',
'eBottom/e5x5',
'eLeft/e5x5',
'eRight/e5x5',
'e2x5Max/e5x5',
'e2x5Left/e5x5',
'e2x5Right/e5x5',
'e2x5Top/e5x5',
'e2x5Bottom/e5x5',
]
EE_vars = common_vars + [
'scPreshowerEnergy/scRawEnergy',
'preshowerEnergyPlane1/scRawEnergy',
'preshowerEnergyPlane2/scRawEnergy',
'iXCoordinate',
'iYCoordinate',
]
EB_vars = common_vars + [
'e5x5/scSeedRawEnergy',
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
print 'Using the following branches for EE:'
print ' ' + '\n '.join(EE_vars)
print 'Using the following branches for EB:'
print ' ' + '\n '.join(EB_vars)
# Write to class
#config.Tree = tree_gDirectory
config.Tree = tree_gDirectory
config.VariablesEE = EE_vars
config.VariablesEB = EB_vars
print "\nAll branches in root file:"
Read_branches_from_rootfile(physical_path(root_file), tree_gDirectory)
########################################
# Output config file
########################################
out_filename = 'photon_config.config'
config.Parse(out_filename)
# Test if the config file can be read by ROOT TEnv
print '\nReading in {0} and trying ROOT.TEnv( ..., 0 ):'.format(
out_filename)
I_TEnv = ROOT.TEnv()
I_TEnv.ReadFile(out_filename, 0)
I_TEnv.Print()
print 'Exited normally'
print '=' * 70
print
def main():
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = "photonConfig"
# filename of the input root file
#root_file = 'FlatNtupFull_13May_SepTrees.root'
#root_file = 'FlatNtupFull_18May_DoubleElectron.root'
#root_file = 'Ntup_20May_DoubleElectron.root'
#root_file = 'Ntup_30May_DoublePhoton_somefailed.root'
#root_file = 'Ntup_01June_DoublePhoton.root'
#root_file = 'Ntup_05June_photons_LowHighPt.root'
# root_file = 'Ntup_12June_photons_lowhighpt.root'
# root_file = 'Ntup_Jun22_lowpt_training.root'
root_file = 'Ntup_Jun22_fullpt_training.root'
# out_filename = 'Config_24Jun_photon_lowpt_OLDVARS.config'
out_filename = 'Config_24Jun_photon_fullpt_OLDVARS.config'
config.Name = out_filename.replace('.config','')
ntup_path = os.path.abspath('../../NTuples/')
# For iterating:
#root_file = 'output.root'
#ntup_path = '/afs/cern.ch/work/t/tklijnsm/EGM/CMSSW_8_0_4/src/SimpleFlatTreeProducer/SimpleNtuplizer/cfgs/'
if not os.path.isdir( ntup_path ):
print 'Error: "{0}"" is not a directory'.format( ntup_path )
physical_path = lambda input_root_file: os.path.join( ntup_path, input_root_file )
########################################
# BDT settings
########################################
config.InputFiles = physical_path( root_file )
# Don't run the Ep combination for photons
config.DoCombine = "False"
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=min(1,exp(-(genPt-50)/50))",
]
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
#config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutBase = "eventNumber%2==0 && genPt<2000"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
config.CutError = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4==3)"
#config.CutComb = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4!=3)"
# # Add an additional cut so that the regression is fast
# NtupIDcut = 200
# config.CutBase += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutError += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutComb += ' && (NtupID<={0})'.format( NtupIDcut )
########################################
# Order Electron tree branches
########################################
# Try to read tree branches from the input root file
#tree_gDirectory = 'een_analyzer/ElectronTree'
tree_gDirectory = 'een_analyzer/PhotonTree'
# ---------------------
# JBs branches for EE (order is important)
# ---------------------
common_vars = [
# Same as in electron case
'nVtx',
'scRawEnergy',
# 'scEta',
# 'scPhi',
'scEtaWidth',
'scPhiWidth',
'scSeedR9',
'scSeedRawEnergy/scRawEnergy',
# 'scSeedLeftRightAsym',
# 'scSeedTopBottomAsym',
'scSeedSigmaIetaIeta',
'scSeedSigmaIetaIphi',
'scSeedSigmaIphiIphi',
'N_ECALClusters',
# 'clusterMaxDR',
# 'clusterMaxDRDPhi',
# 'clusterMaxDRDEta',
# 'clusterMaxDRRawEnergy/scRawEnergy',
# 'clusterRawEnergy[0]/scRawEnergy',
# 'clusterRawEnergy[1]/scRawEnergy',
# 'clusterRawEnergy[2]/scRawEnergy',
# 'clusterDPhiToSeed[0]',
# 'clusterDPhiToSeed[1]',
# 'clusterDPhiToSeed[2]',
# 'clusterDEtaToSeed[0]',
# 'clusterDEtaToSeed[1]',
# 'clusterDEtaToSeed[2]',
# Only for photons
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
'e3x3/e5x5',
'eMax/e5x5',
'e2nd/e5x5',
'eTop/e5x5',
'eBottom/e5x5',
'eLeft/e5x5',
'eRight/e5x5',
'e2x5Max/e5x5',
'e2x5Left/e5x5',
'e2x5Right/e5x5',
'e2x5Top/e5x5',
'e2x5Bottom/e5x5',
]
EE_vars = common_vars + [
'scPreshowerEnergy/scRawEnergy',
'preshowerEnergyPlane1/scRawEnergy',
'preshowerEnergyPlane2/scRawEnergy',
'iXCoordinate',
'iYCoordinate',
]
EB_vars = common_vars + [
'e5x5/scSeedRawEnergy',
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
print 'Using the following branches for EE:'
print ' ' + '\n '.join( EE_vars )
print 'Using the following branches for EB:'
print ' ' + '\n '.join( EB_vars )
# Write to class
#config.Tree = tree_gDirectory
config.Tree = tree_gDirectory
config.VariablesEE = EE_vars
config.VariablesEB = EB_vars
print "\nAll branches in root file:"
Read_branches_from_rootfile( physical_path(root_file) , tree_gDirectory )
########################################
# Output config file
########################################
# out_filename = 'photon_config.config'
# out_filename = 'Config_24Jun_photons_OLDVARS.config'
config.Parse( out_filename )
# Test if the config file can be read by ROOT TEnv
print '\nReading in {0} and trying ROOT.TEnv( ..., 0 ):'.format( out_filename )
I_TEnv = ROOT.TEnv()
I_TEnv.ReadFile( out_filename, 0 )
I_TEnv.Print()
print 'Exited normally'
print '='*70
print
def Make_conf(Verbose=True):
parser = argparse.ArgumentParser()
parser.add_argument(
'--inputrootfile',
'-i',
type=str,
help='Path to root file',
# default='/afs/cern.ch/work/t/tklijnsm/public/CMSSW_8_0_4/src/NTuples/Ntup_Jul22_fullpt_training.root'
default=
'/afs/cern.ch/work/r/rcoelhol/public/CMSSW_8_0_12/src/NTuples/Ntup_10Nov_ElectronPhoton.root'
)
parser.add_argument('--particle',
metavar='N',
type=str,
nargs='+',
help='Specify particles',
default=['electron', 'photon'],
choices=['electron', 'photon'])
parser.add_argument('--region',
metavar='N',
type=str,
nargs='+',
help='Specify regions',
default=['EB', 'EE'],
choices=['EE', 'EB'])
parser.add_argument(
'-n',
'--name',
type=str,
default='NONE',
help='Append a string at the end of the name of this config')
parser.add_argument(
'--fast',
action='store_true',
help='Change some BDT options to be faster (but maybe less precise)')
args = parser.parse_args()
datestr = strftime('%b%d')
return_configs = []
for region in args.region:
for particle in args.particle:
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = 'Config_' + datestr + '_' + particle + '_' + region
if args.name and args.name != 'NONE':
config.Name += '_' + args.name
config.InputFiles = os.path.abspath(args.inputrootfile)
config.Tree = 'een_analyzer/{0}Tree'.format(particle.capitalize())
########################################
# BDT settings
########################################
if args.fast:
config.Options = [
"MinEvents=300",
"Shrinkage=0.15",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=1",
]
config.Name += '_FastOptions'
else:
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=1",
]
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
# Probably not needed
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
if region == 'EB':
config.DoEB = "True"
else:
config.DoEB = "False"
# ======================================
# Sample division - need a part for the ECAL-only training, and a part for the combination
config.CutBase = '1.0'
# These are for the old (regular BDT) EP combination - no longer needed
config.CutComb = '1.0'
config.CutError = '1.0'
# Cut events (otherwise running into CPU limits)
config.CutBase += " && NtupID<4000"
config.CutComb += " && NtupID<4000"
config.CutError += " && NtupID<4000"
########################################
# Order tree branches
########################################
# Agreed list on November 23:
# eval[0] = raw_energy;
# eval[1] = the_sc->etaWidth();
# eval[2] = the_sc->phiWidth();
# eval[3] = full5x5_ess.e5x5/raw_energy;
# eval[4] = ele.hcalOverEcalBc();
# eval[5] = rhoValue_;
# eval[6] = theseed->eta() - the_sc->position().Eta();
# eval[7] = reco::deltaPhi( theseed->phi(),the_sc->position().Phi());
# eval[8] = full5x5_ess.r9;
# eval[9] = full5x5_ess.sigmaIetaIeta;
# eval[10] = full5x5_ess.sigmaIetaIphi;
# eval[11] = full5x5_ess.sigmaIphiIphi;
# eval[12] = full5x5_ess.eMax/full5x5_ess.e5x5;
# eval[13] = full5x5_ess.e2nd/full5x5_ess.e5x5;
# eval[14] = full5x5_ess.eTop/full5x5_ess.e5x5;
# eval[15] = full5x5_ess.eBottom/full5x5_ess.e5x5;
# eval[16] = full5x5_ess.eLeft/full5x5_ess.e5x5;
# eval[17] = full5x5_ess.eRight/full5x5_ess.e5x5;
# eval[18] = EcalClusterToolsT<true>::e2x5Max(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[19] = EcalClusterToolsT<true>::e2x5Left(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[20] = EcalClusterToolsT<true>::e2x5Right(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[21] = EcalClusterToolsT<true>::e2x5Top(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[22] = EcalClusterToolsT<true>::e2x5Bottom(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[23] = N_SATURATEDXTALS;
# eval[24] = std::max(0,numberOfClusters);
# eval[25] = clusterRawEnergy[0]/raw_energy;
# eval[26] = clusterRawEnergy[1]/raw_energy;
# eval[27] = clusterRawEnergy[2]/raw_energy;
# eval[28] = clusterDPhiToSeed[0];
# eval[29] = clusterDPhiToSeed[1];
# eval[30] = clusterDPhiToSeed[2];
# eval[31] = clusterDEtaToSeed[0];
# eval[32] = clusterDEtaToSeed[1];
# eval[33] = clusterDEtaToSeed[2];
# eval[34] = ieta;
# eval[35] = iphi;
# eval[36] = (ieta-signieta)%5;
# eval[37] = (iphi-1)%2;
# eval[38] = (abs(ieta)<=25)*((ieta-signieta)) + (abs(ieta)>25)*((ieta-26*signieta)%20);
# eval[39] = (iphi-1)%20;
# eval[34] = raw_es_energy/raw_energy;
# eval[35] = the_sc->preshowerEnergyPlane1()/raw_energy;
# eval[36] = the_sc->preshowerEnergyPlane2()/raw_energy;
# eval[37] = eeseedid.ix();
# eval[38] = eeseedid.iy();
common_vars = [
# ======================================
# Common variables
'scRawEnergy',
'scEtaWidth',
'scPhiWidth',
'full5x5_e5x5/scRawEnergy',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
# ======================================
# Showershape variables
'full5x5_r9',
'full5x5_sigmaIetaIeta',
'full5x5_sigmaIetaIphi',
'full5x5_sigmaIphiIphi',
'full5x5_eMax/full5x5_e5x5',
'full5x5_e2nd/full5x5_e5x5',
'full5x5_eTop/full5x5_e5x5',
'full5x5_eBottom/full5x5_e5x5',
'full5x5_eLeft/full5x5_e5x5',
'full5x5_eRight/full5x5_e5x5',
'full5x5_e2x5Max/full5x5_e5x5',
'full5x5_e2x5Left/full5x5_e5x5',
'full5x5_e2x5Right/full5x5_e5x5',
'full5x5_e2x5Top/full5x5_e5x5',
'full5x5_e2x5Bottom/full5x5_e5x5',
# ======================================
# Saturation variables
'N_SATURATEDXTALS',
# ======================================
# Cluster variables
'N_ECALClusters',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
# EB specific
config.VariablesEB = common_vars + [
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
# EE specific
config.VariablesEE = common_vars + [
'iXCoordinate',
'iYCoordinate',
'scPreshowerEnergy/scRawEnergy',
# 'preshowerEnergyPlane1/scRawEnergy', # Disabled as of November 2016 (did not influence regression)
# 'preshowerEnergyPlane2/scRawEnergy',# Disabled as of November 2016 (did not influence regression)
]
if Verbose:
print '\n' + '-' * 70
print 'Making config file ' + config.Name + '.config'
print ' Using the following branches for EE:'
print ' ' + '\n '.join(config.VariablesEE)
print ' Using the following branches for EB:'
print ' ' + '\n '.join(config.VariablesEB)
########################################
# Ep combination
########################################
# NOVEMBER 25: NO LONGER NECESSARY TO RUN OLD EP COMBO
config.DoCombine = "False"
config.DoErrors = "False"
########################################
# Output
########################################
# if Verbose:
# # Print all branches as a check
# print "\nAll branches in root file:"
# Read_branches_from_rootfile( physical_path(root_file) , config.Tree )
config.Parse()
# # Test if the config file can be read by ROOT TEnv
# print '\nReading in {0} and trying ROOT.TEnv( ..., 0 ):'.format( out_filename )
# I_TEnv = ROOT.TEnv()
# I_TEnv.ReadFile( out_filename, 0 )
# I_TEnv.Print()
# print 'Exited normally'
# print '='*70
# print
return_configs.append(config)
return return_configs
def Make_conf(Verbose=True):
# root_file = 'Ntup_Jul15_fullpt_training.root'
# Small testing samples -- do NOT use these for plots!
# root_file = 'Ntup_Jun22_lowpt_testing_sample.root'
# Low + high pt sample
# root_file = 'Ntup_Jun22_fullpt_training.root'
# Only low pt sample
# root_file = 'Ntup_Jun22_lowpt_training.root'
# ------------------------------
# 22 July samples - Latest set of branches
# root_file = 'Ntup_Jul22_fullpt_testing_sample.root' # ONLY FOR QUICK TESTS
root_file = 'Ntup_Jul22_fullpt_training.root'
# ntup_path = os.path.join( '/data/userdata/rclsa/ElectronTrees/Jul17/' )
# ntup_path = os.path.join( os.environ['CMSSW_BASE'], 'src/NTuples' )
if os.environ['HOSTNAME'] == 't3ui17':
ntup_path = os.path.join(
'/mnt/t3nfs01/data01/shome/tklijnsm/Samples/RegressionSamples',
'22Jul_samples')
else:
ntup_path = '/afs/cern.ch/work/t/tklijnsm/public/CMSSW_8_0_4/src/NTuples'
datestr = strftime('%b%d')
if not os.path.isdir(ntup_path):
print 'Error: "{0}"" is not a directory'.format(ntup_path)
physical_path = lambda input_root_file: os.path.join(
ntup_path, input_root_file)
return_configs = []
for region in ['EB', 'EE']:
# MODIFIED VERSION
# - Only ECAL_AND_TRK configs made
# - Target set to unweighted
# for ECAL_AND_TRK in [ False, True ]:
for ECAL_AND_TRK in [True]:
for particle in ['electron', 'photon']:
if ECAL_AND_TRK and particle == 'photon':
continue # Photon doesn't have TRK vars
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = 'Config_' + datestr + '_' + particle + '_' + region
# MODIFIED: Force SAMETGT name
# if ECAL_AND_TRK:
# config.Name += '_ECALTRK'
# else:
# config.Name += '_ECALonly'
config.Name += '_SAMETGT'
config.InputFiles = physical_path(root_file)
config.Tree = 'een_analyzer/{0}Tree'.format(
particle.capitalize())
########################################
# BDT settings
########################################
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
# "NTrees=2000", # <-- Moved up from 1000 to include extra tracker effects
"NTrees=1000",
"MinSignificance=5.0",
# "EventWeight=max( min(1,exp(-(genPt-50)/50)), 0.1 )", # <-- What to do?
"EventWeight=1", # <-- No one really likes the weights
]
# MODIFIED: Just force the normal target
# # Set the target - be careful to include the tracker energy in the target for the Ep combination
# if ECAL_AND_TRK:
# config.Target = "genEnergy * (ECALweight + TRKweight) / ( scRawEnergy*ECALweight + scPreshowerEnergy*ECALweight + trkMomentum*TRKweight )"
# else:
# config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
# # config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
if region == 'EB':
config.DoEB = "True"
else:
config.DoEB = "False"
# ======================================
# Sample division - need a part for the ECAL-only training, and a part for the combination
# 80% for the main BDT - divide the sample in divideNumber pieces, and use all but one piece for the main BDT
divideNumber = 3
config.CutBase = "eventNumber%{0}!=0".format(divideNumber)
# 10% for combination, 10% for error
config.CutComb = "eventNumber%{0}==0 && eventNumber%{1}==0".format(
divideNumber, 2 * divideNumber)
config.CutError = "eventNumber%{0}==0 && eventNumber%{1}!=0".format(
divideNumber, 2 * divideNumber)
config.CutBase += " && NtupID<5000"
config.CutComb += " && NtupID<5000"
config.CutError += " && NtupID<5000"
########################################
# Order tree branches
########################################
common_vars = [
# ======================================
# Common variables
# 'pt', # RCLSA: you cannot use the result of the previous training for the new one
# 'nVtx', # rho should be enough information for the BDT
# 'scEta', # Requires alignment information; use crystal number of the seed instead
# 'scPhi', # Requires alignment information; use crystal number of the seed instead
# 'scSeedRawEnergy/scRawEnergy', # RCLSA: Redundant with the one below
'scRawEnergy',
'scEtaWidth',
'scPhiWidth',
'full5x5_e5x5/scRawEnergy',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
# ======================================
# Showershape variables
# Use full 5x5 instead
# 'r9',
# 'eHorizontal',
# 'eVertical',
# 'sigmaIetaIeta',
# 'sigmaIetaIphi',
# 'sigmaIphiIphi',
# 'e5x5',
# 'e3x3',
# 'eMax',
# 'e2nd',
# 'eTop',
# 'eBottom',
# 'eLeft',
# 'eRight',
# 'e2x5Max',
# 'e2x5Left',
# 'e2x5Right',
# 'e2x5Top',
# 'e2x5Bottom',
# Normalization to scRawEnergy necessary?
'full5x5_r9',
# 'full5x5_eHorizontal', # RCLSA: Redundant
# 'full5x5_eVertical', # RCLSA: Redundant
'full5x5_sigmaIetaIeta',
'full5x5_sigmaIetaIphi',
'full5x5_sigmaIphiIphi',
# 'full5x5_e5x5', # RCLSA: Use ratios
# 'full5x5_e3x3/full5x5_e5x5', # RCLSA: Redundant, this is R9
'full5x5_eMax/full5x5_e5x5',
'full5x5_e2nd/full5x5_e5x5',
'full5x5_eTop/full5x5_e5x5',
'full5x5_eBottom/full5x5_e5x5',
'full5x5_eLeft/full5x5_e5x5',
'full5x5_eRight/full5x5_e5x5',
'full5x5_e2x5Max/full5x5_e5x5',
'full5x5_e2x5Left/full5x5_e5x5',
'full5x5_e2x5Right/full5x5_e5x5',
'full5x5_e2x5Top/full5x5_e5x5',
'full5x5_e2x5Bottom/full5x5_e5x5',
# ======================================
# Saturation variables
'N_SATURATEDXTALS',
# 'seedIsSaturated', # RCLSA: probably overkill
# 'seedCrystalEnergy/scSeedRawEnergy', # RCLSA: There is only 1/1e6 cases in which the max energy is not the seed
# ======================================
# Cluster variables
'N_ECALClusters',
# 'clusterMaxDR', # RCLSA Very mismodelled variables
# 'clusterMaxDRDPhi',
# 'clusterMaxDRDEta',
# 'clusterMaxDRRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
if ECAL_AND_TRK:
# ADD THE TRK VARIABLES TO THE MAIN BDT
# Output should be compared to ECAL-only BDT output
common_vars += [
# '( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
# This is simply the corrected energy, I don't we need to pass this (The BDT already has the target in there)
# 'BDTerror/BDTresponse',
# I guess the error we also don't have to pass -- The BDT will already be conscious of the error
'trkMomentumRelError',
# Replace this simply by the trkMomentum only
# 'trkMomentum/(( scRawEnergy + scPreshowerEnergy )*BDTresponse)',
'trkMomentum',
'eleEcalDriven',
# 'full5x5_r9',
'fbrem',
'gsfchi2',
'gsfndof',
'trkEta',
'trkPhi'
]
config.VariablesEB = common_vars + [
# 'cryEtaCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryPhiCoordinate', # Requires alignment information; use crystal number of the seed instead
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
config.VariablesEE = common_vars + [
# 'cryXCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryYCoordinate', # Requires alignment information; use crystal number of the seed instead
'iXCoordinate',
'iYCoordinate',
'scPreshowerEnergy/scRawEnergy',
'preshowerEnergyPlane1/scRawEnergy',
'preshowerEnergyPlane2/scRawEnergy',
]
if Verbose:
print '\n' + '-' * 70
print 'Making config file ' + config.Name + '.config'
print ' Using the following branches for EE:'
print ' ' + '\n '.join(config.VariablesEE)
print ' Using the following branches for EB:'
print ' ' + '\n '.join(config.VariablesEB)
########################################
# Ep combination
########################################
# Only do the combination for the electron AND there are no tracking variables
if particle == 'electron' and not ECAL_AND_TRK:
config.DoCombine = "True"
config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
config.TargetError = "1.253*abs(BDTresponse - genEnergy/(scRawEnergy+scPreshowerEnergy))"
config.VariablesComb = [
'( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
'BDTerror/BDTresponse',
'trkMomentumRelError',
'trkMomentum/(( scRawEnergy + scPreshowerEnergy )*BDTresponse)',
'eleEcalDriven',
'full5x5_r9',
'fbrem',
'gsfchi2',
'gsfndof',
'trkEta',
'trkPhi' # The best way to describe cracks is to use the track (unbiased) directorion
# 'trkMomentum', # RCLSA Again, let us choose one absolute scale and the rest be relative
# 'BDTerror/BDTresponse/trkMomentumRelError',
# ( '( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum *' +
# 'sqrt( BDTerror/BDTresponse*BDTerror/BDTresponse + trkMomentumRelError*trkMomentumRelError)' ),
# 'eleClass',
# 'scIsEB',
]
else:
config.DoCombine = "False"
# # Not necessary if the TRK vars in the main BDT
# config.DoCombine = "False"
########################################
# Output
########################################
# if Verbose:
# # Print all branches as a check
# print "\nAll branches in root file:"
# Read_branches_from_rootfile( physical_path(root_file) , config.Tree )
config.Parse()
# # Test if the config file can be read by ROOT TEnv
# print '\nReading in {0} and trying ROOT.TEnv( ..., 0 ):'.format( out_filename )
# I_TEnv = ROOT.TEnv()
# I_TEnv.ReadFile( out_filename, 0 )
# I_TEnv.Print()
# print 'Exited normally'
# print '='*70
# print
return_configs.append(config)
return return_configs
def Make_conf(Verbose=True):
parser = argparse.ArgumentParser()
parser.add_argument( '--inputrootfile', '-i', type=str, help='Path to root file',
# default='/afs/cern.ch/work/t/tklijnsm/public/CMSSW_8_0_4/src/NTuples/Ntup_Jul22_fullpt_training.root'
default='/afs/cern.ch/work/r/rcoelhol/public/CMSSW_8_0_12/src/NTuples/Ntup_10Nov_ElectronPhoton.root'
)
parser.add_argument(
'--particle', metavar='N', type=str, nargs='+', help='Specify particles',
default=['electron','photon'],choices=['electron','photon']
)
parser.add_argument(
'--region', metavar='N', type=str, nargs='+', help='Specify regions',
default=['EB','EE'],choices=['EE','EB']
)
parser.add_argument(
'-n', '--name', type=str, default='NONE', help='Append a string at the end of the name of this config'
)
parser.add_argument( '--fast', action='store_true', help='Change some BDT options to be faster (but maybe less precise)')
args = parser.parse_args()
datestr = strftime( '%b%d' )
return_configs = []
for region in args.region:
for particle in args.particle:
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = 'Config_' + datestr + '_' + particle + '_' + region
if args.name and args.name!='NONE' : config.Name += '_' + args.name
config.InputFiles = os.path.abspath( args.inputrootfile )
config.Tree = 'een_analyzer/{0}Tree'.format( particle.capitalize() )
########################################
# BDT settings
########################################
if args.fast:
config.Options = [
"MinEvents=300",
"Shrinkage=0.15",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=1",
]
config.Name += '_FastOptions'
else:
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=1",
]
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
# Probably not needed
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
if region == 'EB':
config.DoEB = "True"
else:
config.DoEB = "False"
# ======================================
# Sample division - need a part for the ECAL-only training, and a part for the combination
config.CutBase = '1.0'
# These are for the old (regular BDT) EP combination - no longer needed
config.CutComb = '1.0'
config.CutError = '1.0'
# Cut events (otherwise running into CPU limits)
config.CutBase += " && NtupID<4000"
config.CutComb += " && NtupID<4000"
config.CutError += " && NtupID<4000"
########################################
# Order tree branches
########################################
# Agreed list on November 23:
# eval[0] = raw_energy;
# eval[1] = the_sc->etaWidth();
# eval[2] = the_sc->phiWidth();
# eval[3] = full5x5_ess.e5x5/raw_energy;
# eval[4] = ele.hcalOverEcalBc();
# eval[5] = rhoValue_;
# eval[6] = theseed->eta() - the_sc->position().Eta();
# eval[7] = reco::deltaPhi( theseed->phi(),the_sc->position().Phi());
# eval[8] = full5x5_ess.r9;
# eval[9] = full5x5_ess.sigmaIetaIeta;
# eval[10] = full5x5_ess.sigmaIetaIphi;
# eval[11] = full5x5_ess.sigmaIphiIphi;
# eval[12] = full5x5_ess.eMax/full5x5_ess.e5x5;
# eval[13] = full5x5_ess.e2nd/full5x5_ess.e5x5;
# eval[14] = full5x5_ess.eTop/full5x5_ess.e5x5;
# eval[15] = full5x5_ess.eBottom/full5x5_ess.e5x5;
# eval[16] = full5x5_ess.eLeft/full5x5_ess.e5x5;
# eval[17] = full5x5_ess.eRight/full5x5_ess.e5x5;
# eval[18] = EcalClusterToolsT<true>::e2x5Max(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[19] = EcalClusterToolsT<true>::e2x5Left(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[20] = EcalClusterToolsT<true>::e2x5Right(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[21] = EcalClusterToolsT<true>::e2x5Top(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[22] = EcalClusterToolsT<true>::e2x5Bottom(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[23] = N_SATURATEDXTALS;
# eval[24] = std::max(0,numberOfClusters);
# eval[25] = clusterRawEnergy[0]/raw_energy;
# eval[26] = clusterRawEnergy[1]/raw_energy;
# eval[27] = clusterRawEnergy[2]/raw_energy;
# eval[28] = clusterDPhiToSeed[0];
# eval[29] = clusterDPhiToSeed[1];
# eval[30] = clusterDPhiToSeed[2];
# eval[31] = clusterDEtaToSeed[0];
# eval[32] = clusterDEtaToSeed[1];
# eval[33] = clusterDEtaToSeed[2];
# eval[34] = ieta;
# eval[35] = iphi;
# eval[36] = (ieta-signieta)%5;
# eval[37] = (iphi-1)%2;
# eval[38] = (abs(ieta)<=25)*((ieta-signieta)) + (abs(ieta)>25)*((ieta-26*signieta)%20);
# eval[39] = (iphi-1)%20;
# eval[34] = raw_es_energy/raw_energy;
# eval[35] = the_sc->preshowerEnergyPlane1()/raw_energy;
# eval[36] = the_sc->preshowerEnergyPlane2()/raw_energy;
# eval[37] = eeseedid.ix();
# eval[38] = eeseedid.iy();
common_vars = [
# ======================================
# Common variables
'scRawEnergy',
'scEtaWidth',
'scPhiWidth',
'full5x5_e5x5/scRawEnergy',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
# ======================================
# Showershape variables
'full5x5_r9',
'full5x5_sigmaIetaIeta',
'full5x5_sigmaIetaIphi',
'full5x5_sigmaIphiIphi',
'full5x5_eMax/full5x5_e5x5',
'full5x5_e2nd/full5x5_e5x5',
'full5x5_eTop/full5x5_e5x5',
'full5x5_eBottom/full5x5_e5x5',
'full5x5_eLeft/full5x5_e5x5',
'full5x5_eRight/full5x5_e5x5',
'full5x5_e2x5Max/full5x5_e5x5',
'full5x5_e2x5Left/full5x5_e5x5',
'full5x5_e2x5Right/full5x5_e5x5',
'full5x5_e2x5Top/full5x5_e5x5',
'full5x5_e2x5Bottom/full5x5_e5x5',
# ======================================
# Saturation variables
'N_SATURATEDXTALS',
# ======================================
# Cluster variables
'N_ECALClusters',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
# EB specific
config.VariablesEB = common_vars + [
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
# EE specific
config.VariablesEE = common_vars + [
'iXCoordinate',
'iYCoordinate',
'scPreshowerEnergy/scRawEnergy',
# 'preshowerEnergyPlane1/scRawEnergy', # Disabled as of November 2016 (did not influence regression)
# 'preshowerEnergyPlane2/scRawEnergy',# Disabled as of November 2016 (did not influence regression)
]
if Verbose:
print '\n' + '-'*70
print 'Making config file ' + config.Name + '.config'
print ' Using the following branches for EE:'
print ' ' + '\n '.join( config.VariablesEE )
print ' Using the following branches for EB:'
print ' ' + '\n '.join( config.VariablesEB )
########################################
# Ep combination
########################################
# NOVEMBER 25: NO LONGER NECESSARY TO RUN OLD EP COMBO
config.DoCombine = "False"
config.DoErrors = "False"
########################################
# Output
########################################
# if Verbose:
# # Print all branches as a check
# print "\nAll branches in root file:"
# Read_branches_from_rootfile( physical_path(root_file) , config.Tree )
config.Parse()
# # Test if the config file can be read by ROOT TEnv
# print '\nReading in {0} and trying ROOT.TEnv( ..., 0 ):'.format( out_filename )
# I_TEnv = ROOT.TEnv()
# I_TEnv.ReadFile( out_filename, 0 )
# I_TEnv.Print()
# print 'Exited normally'
# print '='*70
# print
return_configs.append( config )
return return_configs
def main():
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = 'electronConfig'
# filename of the input root file
#root_file = 'FlatNtupFull_13May_SepTrees.root'
#root_file = 'FlatNtupFull_18May_DoubleElectron.root'
#root_file = 'Ntup_20May_DoubleElectron.root'
#root_file = 'Ntup_30May_DoublePhoton_somefailed.root'
#root_file = 'Ntup_01June_DoubleElectron.root'
root_file = 'Ntup_05June_electrons_LowHighPt.root'
ntup_path = os.path.abspath('../../NTuples/')
# For iterating:
#root_file = 'output.root'
#root_file = 'PhotonTest_01June.root'
#ntup_path = '/afs/cern.ch/work/t/tklijnsm/EGM/CMSSW_8_0_4/src/SimpleFlatTreeProducer/SimpleNtuplizer/cfgs/'
if not os.path.isdir(ntup_path):
print 'Error: "{0}"" is not a directory'.format(ntup_path)
physical_path = lambda input_root_file: os.path.join(
ntup_path, input_root_file)
########################################
# BDT settings
########################################
config.InputFiles = physical_path(root_file)
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=min(1,exp(-(genPt-50)/50))",
]
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutBase = "eventNumber%2==0 && genPt<2000"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
config.CutError = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4==3)"
config.CutComb = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4!=3)"
# Add an additional cut so that the regression is fast
# NtupIDcut = 10000
# config.CutBase += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutError += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutComb += ' && (NtupID<={0})'.format( NtupIDcut )
########################################
# Order Electron tree branches
########################################
# Try to read tree branches from the input root file
tree_gDirectory = 'een_analyzer/ElectronTree'
#tree_gDirectory = 'een_analyzer/PhotonTree'
common_vars = [
'nVtx',
'scRawEnergy',
'scEta',
'scPhi',
'scEtaWidth',
'scPhiWidth',
'scSeedR9',
'scSeedRawEnergy/scRawEnergy',
'scSeedEmax',
'scSeedE2nd',
'scSeedLeftRightAsym',
'scSeedTopBottomAsym',
'scSeedSigmaIetaIeta',
'scSeedSigmaIetaIphi',
'scSeedSigmaIphiIphi',
'N_ECALClusters',
'clusterMaxDR',
'clusterMaxDRDPhi',
'clusterMaxDRDEta',
'clusterMaxDRRawEnergy/scRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
EE_vars = common_vars + [
'scPreshowerEnergy/scRawEnergy',
'scSeedCryIxV2',
'scSeedCryIyV2',
]
EB_vars = common_vars + [
'scSeedCryEta',
'scSeedCryPhi',
'scSeedCryIetaV2',
'scSeedCryIphiV2',
]
print 'Using the following branches for EE:'
print ' ' + '\n '.join(EE_vars)
print 'Using the following branches for EB:'
print ' ' + '\n '.join(EB_vars)
# Write to class
#config.Tree = tree_gDirectory
config.Tree = tree_gDirectory
config.VariablesEE = EE_vars
config.VariablesEB = EB_vars
#config.VariablesComb = Ep_branches
config.VariablesComb = [
'( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
'BDTerror/BDTresponse',
'trkMomentum',
'trkMomentumRelError',
'BDTerror/BDTresponse/trkMomentumRelError',
'( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum',
('( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum *' +
'sqrt( BDTerror/BDTresponse*BDTerror/BDTresponse + trkMomentumRelError*trkMomentumRelError)'
),
# 'ecalDriven',
# 'trackerDrivenSeed',
# 'classification',
'eleEcalDriven',
'eleTrackerDriven',
'eleClass',
'scIsEB',
]
print "\nAll branches in root file:"
Read_branches_from_rootfile(physical_path(root_file), tree_gDirectory)
########################################
# Output config file
########################################
out_filename = 'electron_config.config'
config.Parse(out_filename)
# Test if the config file can be read by ROOT TEnv
print '\nReading in {0} and trying ROOT.TEnv( ..., 0 ):'.format(
out_filename)
I_TEnv = ROOT.TEnv()
I_TEnv.ReadFile(out_filename, 0)
I_TEnv.Print()
print 'Exited normally'
print '=' * 70
print
def main():
# Instantiate the Config class which prints a .config file
config = Config()
# config.Name = 'electronConfig'
# filename of the input root file
#root_file = 'FlatNtupFull_13May_SepTrees.root'
#root_file = 'FlatNtupFull_18May_DoubleElectron.root'
#root_file = 'Ntup_20May_DoubleElectron.root'
#root_file = 'Ntup_30May_DoublePhoton_somefailed.root'
#root_file = 'Ntup_01June_DoubleElectron.root'
# root_file = 'Ntup_05June_electrons_LowHighPt.root'
# root_file = 'Ntup_Jun22_lowpt_training.root'
root_file = 'Ntup_Jun22_fullpt_training.root'
# out_filename = 'Config_24Jun_electron_lowpt_OLDVARS.config'
out_filename = 'Config_24Jun_electron_fullpt_OLDVARS.config'
config.Name = out_filename.replace('.config','')
ntup_path = os.path.abspath('../../NTuples/')
# For iterating:
#root_file = 'output.root'
#root_file = 'PhotonTest_01June.root'
#ntup_path = '/afs/cern.ch/work/t/tklijnsm/EGM/CMSSW_8_0_4/src/SimpleFlatTreeProducer/SimpleNtuplizer/cfgs/'
if not os.path.isdir( ntup_path ):
print 'Error: "{0}"" is not a directory'.format( ntup_path )
physical_path = lambda input_root_file: os.path.join( ntup_path, input_root_file )
########################################
# BDT settings
########################################
config.InputFiles = physical_path( root_file )
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=min(1,exp(-(genPt-50)/50))",
]
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutBase = "eventNumber%2==0 && genPt<2000"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
config.CutError = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4==3)"
config.CutComb = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4!=3)"
# Add an additional cut so that the regression is fast
# NtupIDcut = 10000
# config.CutBase += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutError += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutComb += ' && (NtupID<={0})'.format( NtupIDcut )
########################################
# Order Electron tree branches
########################################
# Try to read tree branches from the input root file
tree_gDirectory = 'een_analyzer/ElectronTree'
#tree_gDirectory = 'een_analyzer/PhotonTree'
common_vars = [
'nVtx',
'scRawEnergy',
'scEta',
'scPhi',
'scEtaWidth',
'scPhiWidth',
'scSeedR9',
'scSeedRawEnergy/scRawEnergy',
'scSeedEmax',
'scSeedE2nd',
'scSeedLeftRightAsym',
'scSeedTopBottomAsym',
'scSeedSigmaIetaIeta',
'scSeedSigmaIetaIphi',
'scSeedSigmaIphiIphi',
'N_ECALClusters',
'clusterMaxDR',
'clusterMaxDRDPhi',
'clusterMaxDRDEta',
'clusterMaxDRRawEnergy/scRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
EE_vars = common_vars + [
'scPreshowerEnergy/scRawEnergy',
'scSeedCryIxV2',
'scSeedCryIyV2',
]
EB_vars = common_vars + [
'scSeedCryEta',
'scSeedCryPhi',
'scSeedCryIetaV2',
'scSeedCryIphiV2',
]
print 'Using the following branches for EE:'
print ' ' + '\n '.join( EE_vars )
print 'Using the following branches for EB:'
print ' ' + '\n '.join( EB_vars )
# Write to class
#config.Tree = tree_gDirectory
config.Tree = tree_gDirectory
config.VariablesEE = EE_vars
config.VariablesEB = EB_vars
#config.VariablesComb = Ep_branches
config.VariablesComb = [
'( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
'BDTerror/BDTresponse',
'trkMomentum',
'trkMomentumRelError',
'BDTerror/BDTresponse/trkMomentumRelError',
'( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum',
( '( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum *' +
'sqrt( BDTerror/BDTresponse*BDTerror/BDTresponse + trkMomentumRelError*trkMomentumRelError)' ),
# 'ecalDriven',
# 'trackerDrivenSeed',
# 'classification',
'eleEcalDriven',
'eleTrackerDriven',
'eleClass',
'scIsEB',
]
print "\nAll branches in root file:"
Read_branches_from_rootfile( physical_path(root_file) , tree_gDirectory )
########################################
# Output config file
########################################
config.Parse( out_filename )
# Test if the config file can be read by ROOT TEnv
print '\nReading in {0} and trying ROOT.TEnv( ..., 0 ):'.format( out_filename )
I_TEnv = ROOT.TEnv()
I_TEnv.ReadFile( out_filename, 0 )
I_TEnv.Print()
print 'Exited normally'
print '='*70
print
def main():
# Small testing samples -- do NOT use these for plots!
# root_file = 'Ntup_Jun22_fullpt_testing_sample.root'
# root_file = 'Ntup_Jun22_lowpt_testing_sample.root'
# Low + high pt sample
root_file = 'Ntup_Jun22_fullpt_training.root'
# Only low pt sample
# root_file = 'Ntup_Jun22_lowpt_training.root'
ntup_path = os.path.join('/data/userdata/rclsa/ElectronTrees/')
datestr = strftime('%b%d')
if not os.path.isdir(ntup_path):
print 'Error: "{0}"" is not a directory'.format(ntup_path)
physical_path = lambda input_root_file: os.path.join(
ntup_path, input_root_file)
for particle in ['electron', 'photon']:
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = 'Config_' + particle + '_' + datestr
config.InputFiles = physical_path(root_file)
config.Tree = 'een_analyzer/{0}Tree'.format(particle.capitalize())
########################################
# BDT settings
########################################
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
# "EventWeight=max( min(1,exp(-(genPt-50)/50)), 0.1 )", # <-- What to do?
"EventWeight=1", # <-- No one really likes the weights
]
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutBase = "eventNumber%2==0"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
config.CutError = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4==3)"
# Add an additional cut so that the regression is fast
# NtupIDcut = 10000
# config.CutBase += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutError += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutComb += ' && (NtupID<={0})'.format( NtupIDcut )
########################################
# Order tree branches
########################################
common_vars = [
# ======================================
# Common variables
# 'pt', # RCLSA: you cannot use the result of the previous training for the new one
# 'nVtx', # rho should be enough information for the BDT
'scRawEnergy',
# 'scEta', # Requires alignment information; use crystal number of the seed instead
# 'scPhi', # Requires alignment information; use crystal number of the seed instead
'scEtaWidth',
'scPhiWidth',
# 'scSeedRawEnergy/scRawEnergy', # RCLSA: Redundant with the one below
'full5x5_e5x5/scRawEnergy',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
# ======================================
# Showershape variables
# Use full 5x5 instead
# 'r9',
# 'eHorizontal',
# 'eVertical',
# 'sigmaIetaIeta',
# 'sigmaIetaIphi',
# 'sigmaIphiIphi',
# 'e5x5',
# 'e3x3',
# 'eMax',
# 'e2nd',
# 'eTop',
# 'eBottom',
# 'eLeft',
# 'eRight',
# 'e2x5Max',
# 'e2x5Left',
# 'e2x5Right',
# 'e2x5Top',
# 'e2x5Bottom',
# Normalization to scRawEnergy necessary?
'full5x5_r9',
# 'full5x5_eHorizontal', # RCLSA: Redundant
# 'full5x5_eVertical', # RCLSA: Redundant
'full5x5_sigmaIetaIeta',
'full5x5_sigmaIetaIphi',
'full5x5_sigmaIphiIphi',
# 'full5x5_e5x5', # RCLSA: Use ratios
# 'full5x5_e3x3/full5x5_e5x5', # RCLSA: Redundant, this is R9
'full5x5_eMax/full5x5_e5x5',
'full5x5_e2nd/full5x5_e5x5',
'full5x5_eTop/full5x5_e5x5',
'full5x5_eBottom/full5x5_e5x5',
'full5x5_eLeft/full5x5_e5x5',
'full5x5_eRight/full5x5_e5x5',
'full5x5_e2x5Max/full5x5_e5x5',
'full5x5_e2x5Left/full5x5_e5x5',
'full5x5_e2x5Right/full5x5_e5x5',
'full5x5_e2x5Top/full5x5_e5x5',
'full5x5_e2x5Bottom/full5x5_e5x5',
# ======================================
# Saturation variables
'N_SATURATEDXTALS',
# 'seedIsSaturated', # RCLSA: probably overkill
# 'seedCrystalEnergy/scSeedRawEnergy', # RCLSA: There is only 1/1e6 cases in which the max energy is not the seed
# ======================================
# Cluster variables
'N_ECALClusters',
# 'clusterMaxDR', # RCLSA Very mismodelled variables
# 'clusterMaxDRDPhi',
# 'clusterMaxDRDEta',
# 'clusterMaxDRRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
config.VariablesEB = common_vars + [
# 'cryEtaCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryPhiCoordinate', # Requires alignment information; use crystal number of the seed instead
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
config.VariablesEE = common_vars + [
# 'cryXCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryYCoordinate', # Requires alignment information; use crystal number of the seed instead
'iXCoordinate',
'iYCoordinate',
'scPreshowerEnergy/scRawEnergy',
'preshowerEnergyPlane1/scRawEnergy',
'preshowerEnergyPlane2/scRawEnergy',
]
print 'Using the following branches for EE:'
print ' ' + '\n '.join(config.VariablesEE)
print 'Using the following branches for EB:'
print ' ' + '\n '.join(config.VariablesEB)
########################################
# Ep combination
########################################
# Only do the combination for the electron
if particle == 'electron':
config.DoCombine = "True"
config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
config.CutComb = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4!=3)"
config.VariablesComb = [
'( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
'BDTerror/BDTresponse',
'trkMomentum',
'trkMomentumRelError',
'BDTerror/BDTresponse/trkMomentumRelError',
'( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum',
('( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum *'
+
'sqrt( BDTerror/BDTresponse*BDTerror/BDTresponse + trkMomentumRelError*trkMomentumRelError)'
),
'eleEcalDriven',
'eleTrackerDriven',
'eleClass',
'scIsEB',
]
else:
config.DoCombine = "False"
########################################
# Output
########################################
# Print all branches as a check
print "\nAll branches in root file:"
Read_branches_from_rootfile(physical_path(root_file), config.Tree)
config.Parse()
def Make_conf(Verbose=True):
parser = argparse.ArgumentParser()
parser.add_argument( '--inputrootfile', '-i', type=str, help='Path to root file',
# default='/afs/cern.ch/work/t/tklijnsm/public/CMSSW_8_0_4/src/NTuples/Ntup_Jul22_fullpt_training.root'
default='/afs/cern.ch/work/t/tklijnsm/public/CMSSW_8_0_4/src/NTuples/Ntup_Jul22_fullpt_training.root'
)
parser.add_argument(
'--particle', metavar='N', type=str, nargs='+', help='Specify particles',
default=['electron','photon'],choices=['electron','photon']
)
parser.add_argument(
'--region', metavar='N', type=str, nargs='+', help='Specify regions',
default=['EB','EE'],choices=['EE','EB']
)
parser.add_argument(
'-n', '--name', type=str, default='NONE', help='Append a string at the end of the name of this config'
)
parser.add_argument( '--fast', action='store_true', help='Change some BDT options to be faster (but maybe less precise)')
args = parser.parse_args()
# root_file = 'Ntup_Jul15_fullpt_training.root'
# Small testing samples -- do NOT use these for plots!
# root_file = 'Ntup_Jun22_lowpt_testing_sample.root'
# Low + high pt sample
# root_file = 'Ntup_Jun22_fullpt_training.root'
# Only low pt sample
# root_file = 'Ntup_Jun22_lowpt_training.root'
# ------------------------------
# 22 July samples - Latest set of branches
# root_file = 'Ntup_Jul22_fullpt_testing_sample.root' # ONLY FOR QUICK TESTS
root_file = 'Ntup_Jul22_fullpt_training.root'
# ntup_path = os.path.join( '/data/userdata/rclsa/ElectronTrees/Jul17/' )
# ntup_path = os.path.join( os.environ['CMSSW_BASE'], 'src/NTuples' )
if os.environ['HOSTNAME'] == 't3ui17':
ntup_path = os.path.join( '/mnt/t3nfs01/data01/shome/tklijnsm/Samples/RegressionSamples', '22Jul_samples' )
else:
ntup_path = '/afs/cern.ch/work/t/tklijnsm/public/CMSSW_8_0_4/src/NTuples'
datestr = strftime( '%b%d' )
if not os.path.isdir( ntup_path ):
print 'Error: "{0}"" is not a directory'.format( ntup_path )
physical_path = lambda input_root_file: os.path.join( ntup_path, input_root_file )
return_configs = []
# for region in [ 'EB', 'EE' ]:
for region in args.region:
for ECAL_AND_TRK in [ False ]:
# for particle in [ 'electron', 'photon' ]:
for particle in args.particle:
if ECAL_AND_TRK and particle=='photon': continue # Photon doesn't have TRK vars
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = 'Config_' + datestr + '_' + particle + '_' + region
if ECAL_AND_TRK:
config.Name += '_ECALTRK'
else:
config.Name += '_ECALonly'
if args.name and args.name!='NONE' : config.Name += '_' + args.name
config.InputFiles = physical_path( root_file )
config.Tree = 'een_analyzer/{0}Tree'.format( particle.capitalize() )
########################################
# BDT settings
########################################
# config.Options = [
# "MinEvents=200",
# "Shrinkage=0.1",
# "NTrees=1000",
# "MinSignificance=5.0",
# "EventWeight=1",
# ]
config.Options = [
"MinEvents=300",
"Shrinkage=0.15",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=1",
]
# Set the target - be careful to include the tracker energy in the target for the Ep combination
if ECAL_AND_TRK:
config.Target = "genEnergy * (ECALweight + TRKweight) / ( scRawEnergy*ECALweight + scPreshowerEnergy*ECALweight + trkMomentum*TRKweight )"
else:
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
# config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
if region == 'EB':
config.DoEB = "True"
else:
config.DoEB = "False"
# ======================================
# Sample division - need a part for the ECAL-only training, and a part for the combination
# WITH EP COMBINATION TRAINING IN SERIES, THIS IS NO LONGER NEEDED
# if particle == 'electron':
# # 66% for the main BDT - divide the sample in divideNumber pieces, and use all but one piece for the main BDT
# divideNumber = 3
# config.CutBase = "eventNumber%{0}!=0".format( divideNumber )
# # 17% for combination, 17% for error
# config.CutComb = "eventNumber%{0}==0 && eventNumber%{1}==0".format( divideNumber, 2*divideNumber )
# config.CutError = "eventNumber%{0}==0 && eventNumber%{1}!=0".format( divideNumber, 2*divideNumber )
# elif particle == 'photon':
# config.CutBase = '1.0'
# config.CutComb = '1.0'
# config.CutError = '1.0'
config.CutBase = '1.0'
config.CutComb = '1.0'
config.CutError = '1.0'
# # TEMPORARY: cut events drastically for test mode
config.CutBase += " && NtupID<4000"
config.CutComb += " && NtupID<4000"
config.CutError += " && NtupID<4000"
########################################
# Order tree branches
########################################
# Agreed list on November 23:
# eval[0] = raw_energy;
# eval[1] = the_sc->etaWidth();
# eval[2] = the_sc->phiWidth();
# eval[3] = full5x5_ess.e5x5/raw_energy;
# eval[4] = ele.hcalOverEcalBc();
# eval[5] = rhoValue_;
# eval[6] = theseed->eta() - the_sc->position().Eta();
# eval[7] = reco::deltaPhi( theseed->phi(),the_sc->position().Phi());
# eval[8] = full5x5_ess.r9;
# eval[9] = full5x5_ess.sigmaIetaIeta;
# eval[10] = full5x5_ess.sigmaIetaIphi;
# eval[11] = full5x5_ess.sigmaIphiIphi;
# eval[12] = full5x5_ess.eMax/full5x5_ess.e5x5;
# eval[13] = full5x5_ess.e2nd/full5x5_ess.e5x5;
# eval[14] = full5x5_ess.eTop/full5x5_ess.e5x5;
# eval[15] = full5x5_ess.eBottom/full5x5_ess.e5x5;
# eval[16] = full5x5_ess.eLeft/full5x5_ess.e5x5;
# eval[17] = full5x5_ess.eRight/full5x5_ess.e5x5;
# eval[18] = EcalClusterToolsT<true>::e2x5Max(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[19] = EcalClusterToolsT<true>::e2x5Left(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[20] = EcalClusterToolsT<true>::e2x5Right(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[21] = EcalClusterToolsT<true>::e2x5Top(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[22] = EcalClusterToolsT<true>::e2x5Bottom(*theseed, &*ecalRecHits, topology_)/full5x5_ess.e5x5;
# eval[23] = N_SATURATEDXTALS;
# eval[24] = std::max(0,numberOfClusters);
# eval[25] = clusterRawEnergy[0]/raw_energy;
# eval[26] = clusterRawEnergy[1]/raw_energy;
# eval[27] = clusterRawEnergy[2]/raw_energy;
# eval[28] = clusterDPhiToSeed[0];
# eval[29] = clusterDPhiToSeed[1];
# eval[30] = clusterDPhiToSeed[2];
# eval[31] = clusterDEtaToSeed[0];
# eval[32] = clusterDEtaToSeed[1];
# eval[33] = clusterDEtaToSeed[2];
# eval[34] = ieta;
# eval[35] = iphi;
# eval[36] = (ieta-signieta)%5;
# eval[37] = (iphi-1)%2;
# eval[38] = (abs(ieta)<=25)*((ieta-signieta)) + (abs(ieta)>25)*((ieta-26*signieta)%20);
# eval[39] = (iphi-1)%20;
# eval[34] = raw_es_energy/raw_energy;
# eval[35] = the_sc->preshowerEnergyPlane1()/raw_energy;
# eval[36] = the_sc->preshowerEnergyPlane2()/raw_energy;
# eval[37] = eeseedid.ix();
# eval[38] = eeseedid.iy();
common_vars = [
# ======================================
# Common variables
# 'pt', # RCLSA: you cannot use the result of the previous training for the new one
# 'nVtx', # rho should be enough information for the BDT
# 'scEta', # Requires alignment information; use crystal number of the seed instead
# 'scPhi', # Requires alignment information; use crystal number of the seed instead
# 'scSeedRawEnergy/scRawEnergy', # RCLSA: Redundant with the one below
'scRawEnergy',
'scEtaWidth',
'scPhiWidth',
'full5x5_e5x5/scRawEnergy',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
# ======================================
# Showershape variables
# Use full 5x5 instead
# 'r9',
# 'eHorizontal',
# 'eVertical',
# 'sigmaIetaIeta',
# 'sigmaIetaIphi',
# 'sigmaIphiIphi',
# 'e5x5',
# 'e3x3',
# 'eMax',
# 'e2nd',
# 'eTop',
# 'eBottom',
# 'eLeft',
# 'eRight',
# 'e2x5Max',
# 'e2x5Left',
# 'e2x5Right',
# 'e2x5Top',
# 'e2x5Bottom',
# Normalization to scRawEnergy necessary?
'full5x5_r9',
# 'full5x5_eHorizontal', # RCLSA: Redundant
# 'full5x5_eVertical', # RCLSA: Redundant
'full5x5_sigmaIetaIeta',
'full5x5_sigmaIetaIphi',
'full5x5_sigmaIphiIphi',
# 'full5x5_e5x5', # RCLSA: Use ratios
# 'full5x5_e3x3/full5x5_e5x5', # RCLSA: Redundant, this is R9
'full5x5_eMax/full5x5_e5x5',
'full5x5_e2nd/full5x5_e5x5',
'full5x5_eTop/full5x5_e5x5',
'full5x5_eBottom/full5x5_e5x5',
'full5x5_eLeft/full5x5_e5x5',
'full5x5_eRight/full5x5_e5x5',
'full5x5_e2x5Max/full5x5_e5x5',
'full5x5_e2x5Left/full5x5_e5x5',
'full5x5_e2x5Right/full5x5_e5x5',
'full5x5_e2x5Top/full5x5_e5x5',
'full5x5_e2x5Bottom/full5x5_e5x5',
# ======================================
# Saturation variables
'N_SATURATEDXTALS',
# 'seedIsSaturated', # RCLSA: probably overkill
# 'seedCrystalEnergy/scSeedRawEnergy', # RCLSA: There is only 1/1e6 cases in which the max energy is not the seed
# ======================================
# Cluster variables
'N_ECALClusters',
# 'clusterMaxDR', # RCLSA Very mismodelled variables
# 'clusterMaxDRDPhi',
# 'clusterMaxDRDEta',
# 'clusterMaxDRRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
if ECAL_AND_TRK:
# ADD THE TRK VARIABLES TO THE MAIN BDT
# Output should be compared to ECAL-only BDT output
common_vars += [
# '( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
# This is simply the corrected energy, I don't we need to pass this (The BDT already has the target in there)
# 'BDTerror/BDTresponse',
# I guess the error we also don't have to pass -- The BDT will already be conscious of the error
'trkMomentumRelError',
# Replace this simply by the trkMomentum only
# 'trkMomentum/(( scRawEnergy + scPreshowerEnergy )*BDTresponse)',
'trkMomentum',
'eleEcalDriven',
# 'full5x5_r9',
'fbrem',
'gsfchi2',
'gsfndof',
'trkEta',
'trkPhi'
]
config.VariablesEB = common_vars + [
# 'cryEtaCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryPhiCoordinate', # Requires alignment information; use crystal number of the seed instead
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
config.VariablesEE = common_vars + [
# 'cryXCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryYCoordinate', # Requires alignment information; use crystal number of the seed instead
'iXCoordinate',
'iYCoordinate',
'scPreshowerEnergy/scRawEnergy',
# 'preshowerEnergyPlane1/scRawEnergy',
# 'preshowerEnergyPlane2/scRawEnergy',
]
if Verbose:
print '\n' + '-'*70
print 'Making config file ' + config.Name + '.config'
print ' Using the following branches for EE:'
print ' ' + '\n '.join( config.VariablesEE )
print ' Using the following branches for EB:'
print ' ' + '\n '.join( config.VariablesEB )
########################################
# Ep combination
########################################
# NOVEMBER 25: NO LONGER NECESSARY TO RUN OLD EP COMBO
config.DoCombine = "False"
# # Only do the combination for the electron AND there are no tracking variables
# if particle == 'electron' and not ECAL_AND_TRK:
# config.DoCombine = "True"
# config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
# config.TargetError = "1.253*abs(BDTresponse - genEnergy/(scRawEnergy+scPreshowerEnergy))"
# config.VariablesComb = [
# '( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
# 'BDTerror/BDTresponse',
# 'trkMomentumRelError',
# 'trkMomentum/(( scRawEnergy + scPreshowerEnergy )*BDTresponse)',
# 'eleEcalDriven',
# 'full5x5_r9',
# 'fbrem',
# 'gsfchi2',
# 'gsfndof',
# 'trkEta',
# 'trkPhi' # The best way to describe cracks is to use the track (unbiased) directorion
# # 'trkMomentum', # RCLSA Again, let us choose one absolute scale and the rest be relative
# # 'BDTerror/BDTresponse/trkMomentumRelError',
# # ( '( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum *' +
# # 'sqrt( BDTerror/BDTresponse*BDTerror/BDTresponse + trkMomentumRelError*trkMomentumRelError)' ),
# # 'eleClass',
# # 'scIsEB',
# ]
# else:
# config.DoCombine = "False"
########################################
# Output
########################################
# if Verbose:
# # Print all branches as a check
# print "\nAll branches in root file:"
# Read_branches_from_rootfile( physical_path(root_file) , config.Tree )
config.Parse()
# # Test if the config file can be read by ROOT TEnv
# print '\nReading in {0} and trying ROOT.TEnv( ..., 0 ):'.format( out_filename )
# I_TEnv = ROOT.TEnv()
# I_TEnv.ReadFile( out_filename, 0 )
# I_TEnv.Print()
# print 'Exited normally'
# print '='*70
# print
return_configs.append( config )
return return_configs
def Make_conf(Verbose=True):
# root_file = 'Ntup_Jul15_fullpt_training.root'
# Small testing samples -- do NOT use these for plots!
# root_file = 'Ntup_Jun22_lowpt_testing_sample.root'
# Low + high pt sample
# root_file = 'Ntup_Jun22_fullpt_training.root'
# Only low pt sample
# root_file = 'Ntup_Jun22_lowpt_training.root'
# ------------------------------
# 22 July samples - Latest set of branches
# root_file = 'Ntup_Jul22_fullpt_testing_sample.root' # ONLY FOR QUICK TESTS
root_file = 'Ntup_Jul22_fullpt_training.root'
# ntup_path = os.path.join( '/data/userdata/rclsa/ElectronTrees/Jul17/' )
# ntup_path = os.path.join( os.environ['CMSSW_BASE'], 'src/NTuples' )
if os.environ['HOSTNAME'] == 't3ui17':
ntup_path = os.path.join( '/mnt/t3nfs01/data01/shome/tklijnsm/Samples/RegressionSamples', '22Jul_samples' )
else:
ntup_path = '/afs/cern.ch/work/t/tklijnsm/public/CMSSW_8_0_4/src/NTuples'
datestr = strftime( '%b%d' )
if not os.path.isdir( ntup_path ):
print 'Error: "{0}"" is not a directory'.format( ntup_path )
physical_path = lambda input_root_file: os.path.join( ntup_path, input_root_file )
return_configs = []
for region in [ 'EB', 'EE' ]:
# MODIFIED VERSION
# - Only ECAL_AND_TRK configs made
# - Target set to unweighted
# for ECAL_AND_TRK in [ False, True ]:
for ECAL_AND_TRK in [ True ]:
for particle in [ 'electron', 'photon' ]:
if ECAL_AND_TRK and particle=='photon': continue # Photon doesn't have TRK vars
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = 'Config_' + datestr + '_' + particle + '_' + region
# MODIFIED: Force SAMETGT name
# if ECAL_AND_TRK:
# config.Name += '_ECALTRK'
# else:
# config.Name += '_ECALonly'
config.Name += '_SAMETGT'
config.InputFiles = physical_path( root_file )
config.Tree = 'een_analyzer/{0}Tree'.format( particle.capitalize() )
########################################
# BDT settings
########################################
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
# "NTrees=2000", # <-- Moved up from 1000 to include extra tracker effects
"NTrees=1000",
"MinSignificance=5.0",
# "EventWeight=max( min(1,exp(-(genPt-50)/50)), 0.1 )", # <-- What to do?
"EventWeight=1", # <-- No one really likes the weights
]
# MODIFIED: Just force the normal target
# # Set the target - be careful to include the tracker energy in the target for the Ep combination
# if ECAL_AND_TRK:
# config.Target = "genEnergy * (ECALweight + TRKweight) / ( scRawEnergy*ECALweight + scPreshowerEnergy*ECALweight + trkMomentum*TRKweight )"
# else:
# config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
# # config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
if region == 'EB':
config.DoEB = "True"
else:
config.DoEB = "False"
# ======================================
# Sample division - need a part for the ECAL-only training, and a part for the combination
# 80% for the main BDT - divide the sample in divideNumber pieces, and use all but one piece for the main BDT
divideNumber = 3
config.CutBase = "eventNumber%{0}!=0".format( divideNumber )
# 10% for combination, 10% for error
config.CutComb = "eventNumber%{0}==0 && eventNumber%{1}==0".format( divideNumber, 2*divideNumber )
config.CutError = "eventNumber%{0}==0 && eventNumber%{1}!=0".format( divideNumber, 2*divideNumber )
config.CutBase += " && NtupID<5000"
config.CutComb += " && NtupID<5000"
config.CutError += " && NtupID<5000"
########################################
# Order tree branches
########################################
common_vars = [
# ======================================
# Common variables
# 'pt', # RCLSA: you cannot use the result of the previous training for the new one
# 'nVtx', # rho should be enough information for the BDT
# 'scEta', # Requires alignment information; use crystal number of the seed instead
# 'scPhi', # Requires alignment information; use crystal number of the seed instead
# 'scSeedRawEnergy/scRawEnergy', # RCLSA: Redundant with the one below
'scRawEnergy',
'scEtaWidth',
'scPhiWidth',
'full5x5_e5x5/scRawEnergy',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
# ======================================
# Showershape variables
# Use full 5x5 instead
# 'r9',
# 'eHorizontal',
# 'eVertical',
# 'sigmaIetaIeta',
# 'sigmaIetaIphi',
# 'sigmaIphiIphi',
# 'e5x5',
# 'e3x3',
# 'eMax',
# 'e2nd',
# 'eTop',
# 'eBottom',
# 'eLeft',
# 'eRight',
# 'e2x5Max',
# 'e2x5Left',
# 'e2x5Right',
# 'e2x5Top',
# 'e2x5Bottom',
# Normalization to scRawEnergy necessary?
'full5x5_r9',
# 'full5x5_eHorizontal', # RCLSA: Redundant
# 'full5x5_eVertical', # RCLSA: Redundant
'full5x5_sigmaIetaIeta',
'full5x5_sigmaIetaIphi',
'full5x5_sigmaIphiIphi',
# 'full5x5_e5x5', # RCLSA: Use ratios
# 'full5x5_e3x3/full5x5_e5x5', # RCLSA: Redundant, this is R9
'full5x5_eMax/full5x5_e5x5',
'full5x5_e2nd/full5x5_e5x5',
'full5x5_eTop/full5x5_e5x5',
'full5x5_eBottom/full5x5_e5x5',
'full5x5_eLeft/full5x5_e5x5',
'full5x5_eRight/full5x5_e5x5',
'full5x5_e2x5Max/full5x5_e5x5',
'full5x5_e2x5Left/full5x5_e5x5',
'full5x5_e2x5Right/full5x5_e5x5',
'full5x5_e2x5Top/full5x5_e5x5',
'full5x5_e2x5Bottom/full5x5_e5x5',
# ======================================
# Saturation variables
'N_SATURATEDXTALS',
# 'seedIsSaturated', # RCLSA: probably overkill
# 'seedCrystalEnergy/scSeedRawEnergy', # RCLSA: There is only 1/1e6 cases in which the max energy is not the seed
# ======================================
# Cluster variables
'N_ECALClusters',
# 'clusterMaxDR', # RCLSA Very mismodelled variables
# 'clusterMaxDRDPhi',
# 'clusterMaxDRDEta',
# 'clusterMaxDRRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
if ECAL_AND_TRK:
# ADD THE TRK VARIABLES TO THE MAIN BDT
# Output should be compared to ECAL-only BDT output
common_vars += [
# '( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
# This is simply the corrected energy, I don't we need to pass this (The BDT already has the target in there)
# 'BDTerror/BDTresponse',
# I guess the error we also don't have to pass -- The BDT will already be conscious of the error
'trkMomentumRelError',
# Replace this simply by the trkMomentum only
# 'trkMomentum/(( scRawEnergy + scPreshowerEnergy )*BDTresponse)',
'trkMomentum',
'eleEcalDriven',
# 'full5x5_r9',
'fbrem',
'gsfchi2',
'gsfndof',
'trkEta',
'trkPhi'
]
config.VariablesEB = common_vars + [
# 'cryEtaCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryPhiCoordinate', # Requires alignment information; use crystal number of the seed instead
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
config.VariablesEE = common_vars + [
# 'cryXCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryYCoordinate', # Requires alignment information; use crystal number of the seed instead
'iXCoordinate',
'iYCoordinate',
'scPreshowerEnergy/scRawEnergy',
'preshowerEnergyPlane1/scRawEnergy',
'preshowerEnergyPlane2/scRawEnergy',
]
if Verbose:
print '\n' + '-'*70
print 'Making config file ' + config.Name + '.config'
print ' Using the following branches for EE:'
print ' ' + '\n '.join( config.VariablesEE )
print ' Using the following branches for EB:'
print ' ' + '\n '.join( config.VariablesEB )
########################################
# Ep combination
########################################
# Only do the combination for the electron AND there are no tracking variables
if particle == 'electron' and not ECAL_AND_TRK:
config.DoCombine = "True"
config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
config.TargetError = "1.253*abs(BDTresponse - genEnergy/(scRawEnergy+scPreshowerEnergy))"
config.VariablesComb = [
'( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
'BDTerror/BDTresponse',
'trkMomentumRelError',
'trkMomentum/(( scRawEnergy + scPreshowerEnergy )*BDTresponse)',
'eleEcalDriven',
'full5x5_r9',
'fbrem',
'gsfchi2',
'gsfndof',
'trkEta',
'trkPhi' # The best way to describe cracks is to use the track (unbiased) directorion
# 'trkMomentum', # RCLSA Again, let us choose one absolute scale and the rest be relative
# 'BDTerror/BDTresponse/trkMomentumRelError',
# ( '( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum *' +
# 'sqrt( BDTerror/BDTresponse*BDTerror/BDTresponse + trkMomentumRelError*trkMomentumRelError)' ),
# 'eleClass',
# 'scIsEB',
]
else:
config.DoCombine = "False"
# # Not necessary if the TRK vars in the main BDT
# config.DoCombine = "False"
########################################
# Output
########################################
# if Verbose:
# # Print all branches as a check
# print "\nAll branches in root file:"
# Read_branches_from_rootfile( physical_path(root_file) , config.Tree )
config.Parse()
# # Test if the config file can be read by ROOT TEnv
# print '\nReading in {0} and trying ROOT.TEnv( ..., 0 ):'.format( out_filename )
# I_TEnv = ROOT.TEnv()
# I_TEnv.ReadFile( out_filename, 0 )
# I_TEnv.Print()
# print 'Exited normally'
# print '='*70
# print
return_configs.append( config )
return return_configs
def main():
# Small testing samples -- do NOT use these for plots!
# fullpt_root_file = 'Ntup_Jun22_fullpt_testing_sample.root'
# lowpt_root_file = 'Ntup_Jun22_lowpt_testing_sample.root'
# Low + high pt sample
fullpt_root_file = 'Ntup_Jun22_fullpt_training.root'
# Only low pt sample
lowpt_root_file = 'Ntup_Jun22_lowpt_training.root'
ntup_path = os.path.join(os.environ['CMSSW_BASE'], 'src/NTuples')
datestr = strftime('%b%d')
if not os.path.isdir(ntup_path):
print 'Error: "{0}"" is not a directory'.format(ntup_path)
physical_path = lambda input_root_file: os.path.join(
ntup_path, input_root_file)
########################################
# BASE CONFIG - This is low pt electrons
# Configs for photons and and other pt ranges are created by altering this one
########################################
# Instantiate the Config class which prints a .config file
base_config = Config()
base_config.Name = 'Config_electron_lowpt_' + datestr
base_config.InputFiles = physical_path(lowpt_root_file)
base_config.Tree = 'een_analyzer/ElectronTree'
########################################
# BDT settings
########################################
base_config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=max( min(1,exp(-(genPt-50)/50)), 0.1 )", # <-- What to do?
]
base_config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
base_config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
base_config.HistoConfig = "jobs/dummy_Histo.config"
base_config.CutBase = "eventNumber%2==0"
base_config.CutEB = "scIsEB"
base_config.CutEE = "!scIsEB"
base_config.CutError = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4==3)"
# Add an additional cut so that the regression is fast
# NtupIDcut = 10000
# base_config.CutBase += ' && (NtupID<={0})'.format( NtupIDcut )
# base_config.CutError += ' && (NtupID<={0})'.format( NtupIDcut )
# base_config.CutComb += ' && (NtupID<={0})'.format( NtupIDcut )
########################################
# Order tree branches
########################################
common_vars = [
# ======================================
# Common variables
'pt',
# 'nVtx', # rho should be enough information for the BDT
'scRawEnergy',
# 'scEta', # Requires alignment information; use crystal number of the seed instead
# 'scPhi', # Requires alignment information; use crystal number of the seed instead
'scEtaWidth',
'scPhiWidth',
'scSeedRawEnergy/scRawEnergy',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
# ======================================
# Showershape variables
# Use full 5x5 instead
# 'r9',
# 'eHorizontal',
# 'eVertical',
# 'sigmaIetaIeta',
# 'sigmaIetaIphi',
# 'sigmaIphiIphi',
# 'e5x5',
# 'e3x3',
# 'eMax',
# 'e2nd',
# 'eTop',
# 'eBottom',
# 'eLeft',
# 'eRight',
# 'e2x5Max',
# 'e2x5Left',
# 'e2x5Right',
# 'e2x5Top',
# 'e2x5Bottom',
# Normalization to scRawEnergy necessary?
'full5x5_r9',
'full5x5_eHorizontal',
'full5x5_eVertical',
'full5x5_sigmaIetaIeta',
'full5x5_sigmaIetaIphi',
'full5x5_sigmaIphiIphi',
'full5x5_e5x5',
'full5x5_e3x3',
'full5x5_eMax',
'full5x5_e2nd',
'full5x5_eTop',
'full5x5_eBottom',
'full5x5_eLeft',
'full5x5_eRight',
'full5x5_e2x5Max',
'full5x5_e2x5Left',
'full5x5_e2x5Right',
'full5x5_e2x5Top',
'full5x5_e2x5Bottom',
# ======================================
# Saturation variables
'N_SATURATEDXTALS',
'seedIsSaturated',
'seedCrystalEnergy/scRawEnergy',
# ======================================
# Cluster variables
'N_ECALClusters',
'clusterMaxDR',
'clusterMaxDRDPhi',
'clusterMaxDRDEta',
'clusterMaxDRRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
base_config.VariablesEB = common_vars + [
# 'cryEtaCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryPhiCoordinate', # Requires alignment information; use crystal number of the seed instead
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
base_config.VariablesEE = common_vars + [
# 'cryXCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryYCoordinate', # Requires alignment information; use crystal number of the seed instead
'iXCoordinate',
'iYCoordinate',
'scPreshowerEnergy/scRawEnergy',
'preshowerEnergyPlane1/scRawEnergy',
'preshowerEnergyPlane2/scRawEnergy',
]
# print 'Using the following branches for EE:'
# print ' ' + '\n '.join( base_config.VariablesEE )
# print 'Using the following branches for EB:'
# print ' ' + '\n '.join( base_config.VariablesEB )
########################################
# Ep combination
########################################
# Only do the combination for the electron
base_config.DoCombine = "True"
base_config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
base_config.CutComb = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4!=3)"
base_config.VariablesComb = [
'( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
'BDTerror/BDTresponse',
'trkMomentum',
'trkMomentumRelError',
'BDTerror/BDTresponse/trkMomentumRelError',
'( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum',
('( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum *' +
'sqrt( BDTerror/BDTresponse*BDTerror/BDTresponse + trkMomentumRelError*trkMomentumRelError)'
),
'eleEcalDriven',
'eleTrackerDriven',
'eleClass',
'scIsEB',
]
########################################
# Output
########################################
# lowpt electrons - this is simply the base config defined above
base_config.Parse()
# fullpt electrons - only change the root file
base_config.Name = 'Config_electron_fullpt_' + datestr
base_config.InputFiles = physical_path(fullpt_root_file)
base_config.Parse()
# lowpt photons
base_config.Name = 'Config_photon_lowpt_' + datestr
base_config.InputFiles = physical_path(lowpt_root_file)
base_config.Tree = 'een_analyzer/PhotonTree'
base_config.DoCombine = "False"
base_config.Parse()
# fullpt photons
base_config.Name = 'Config_photon_fullpt_' + datestr
base_config.InputFiles = physical_path(fullpt_root_file)
base_config.Tree = 'een_analyzer/PhotonTree'
base_config.DoCombine = "False"
base_config.Parse()
########################################
# OLD VARIABLES
########################################
# Remove the max( ..., 0.1, ) from the eventweight
base_config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=min(1,exp(-(genPt-50)/50))",
]
# lowpt electrons
base_config.Name = 'Config_electron_lowpt_' + datestr + '_OLDVARS'
base_config.InputFiles = physical_path(lowpt_root_file)
base_config.Tree = 'een_analyzer/ElectronTree'
base_config.DoCombine = "True"
OLD_common_electron_vars = [
'nVtx',
'scRawEnergy',
'scEta',
'scPhi',
'scEtaWidth',
'scPhiWidth',
'r9',
'scSeedRawEnergy/scRawEnergy',
'eMax',
'e2nd',
'eHorizontal', # 'scSeedLeftRightAsym',
'eVertical', # 'scSeedTopBottomAsym',
'sigmaIetaIeta',
'sigmaIetaIphi',
'sigmaIphiIphi',
'N_ECALClusters',
'clusterMaxDR',
'clusterMaxDRDPhi',
'clusterMaxDRDEta',
'clusterMaxDRRawEnergy/scRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
base_config.VariablesEB = OLD_common_electron_vars + [
'cryEtaCoordinate',
'cryPhiCoordinate',
'iEtaCoordinate',
'iPhiCoordinate',
# 'scSeedCryEta',
# 'scSeedCryPhi',
# 'scSeedCryIetaV2',
# 'scSeedCryIphiV2',
]
base_config.VariablesEE = OLD_common_electron_vars + [
'scPreshowerEnergy/scRawEnergy',
# 'scSeedCryIxV2',
# 'scSeedCryIyV2',
'iXCoordinate',
'iYCoordinate',
]
base_config.Parse()
# fullpt oldvars
base_config.Name = 'Config_electron_fullpt_' + datestr + '_OLDVARS'
base_config.InputFiles = physical_path(fullpt_root_file)
base_config.Parse()
# lowpt photons
base_config.Name = 'Config_photon_lowpt_' + datestr + '_OLDVARS'
base_config.InputFiles = physical_path(lowpt_root_file)
base_config.Tree = 'een_analyzer/PhotonTree'
base_config.DoCombine = "False"
OLD_common_photon_vars = [
'nVtx',
'scRawEnergy',
# 'scEta',
# 'scPhi',
'scEtaWidth',
'scPhiWidth',
'r9',
'scSeedRawEnergy/scRawEnergy',
# 'scSeedLeftRightAsym',
# 'scSeedTopBottomAsym',
'sigmaIetaIeta',
'sigmaIetaIphi',
'sigmaIphiIphi',
'N_ECALClusters',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
'e3x3/e5x5',
'eMax/e5x5',
'e2nd/e5x5',
'eTop/e5x5',
'eBottom/e5x5',
'eLeft/e5x5',
'eRight/e5x5',
'e2x5Max/e5x5',
'e2x5Left/e5x5',
'e2x5Right/e5x5',
'e2x5Top/e5x5',
'e2x5Bottom/e5x5',
]
base_config.VariablesEB = OLD_common_photon_vars + [
'e5x5/scSeedRawEnergy',
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
base_config.VariablesEE = OLD_common_photon_vars + [
'scPreshowerEnergy/scRawEnergy',
'preshowerEnergyPlane1/scRawEnergy',
'preshowerEnergyPlane2/scRawEnergy',
'iXCoordinate',
'iYCoordinate',
]
base_config.Parse()
# fullpt photons
base_config.Name = 'Config_photon_fullpt_' + datestr + '_OLDVARS'
base_config.InputFiles = physical_path(fullpt_root_file)
base_config.Parse()
# Print all branches as a check
print "\nAll branches in lowpt root file:"
Read_branches_from_rootfile(physical_path(lowpt_root_file),
base_config.Tree)
def MakeConf(Verbose=True):
parser = argparse.ArgumentParser()
parser.add_argument( '--inputrootfile', '-i', type=str, help='Path to root file',
# default='../applyRegression/Config_Sep30_electron_EB_ECALonly_appliedRegression_training.root'
# default='../Config_Oct25_electron_EB_ECALonly_appliedRegression_training_ptWeight.root'
)
parser.add_argument(
'--region', metavar='N', type=str, nargs='+', help='Specify regions',
default=['EB','EE'],choices=['EE','EB']
)
parser.add_argument(
'-n', '--name', type=str, default='NONE', help='Append a string at the end of the name of this config'
)
parser.add_argument( '--fast', action='store_true', help='Change some BDT options to be faster (but maybe less precise)')
args = parser.parse_args()
datestr = strftime( '%b%d' )
# Photon does not have TRK vars
particle = 'electron'
# Reads off the name of this .py file, so it's clear what made this.
moduleName = os.path.basename(__file__).replace('MakeConf_','').replace('.py','')
return_configs = []
for region in args.region:
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = 'Config_' + datestr + '_' + particle + '_' + region + '_' + moduleName
# Append a string to the name if given by the user
if not args.name == 'NONE':
config.Name += '_' + args.name
if args.fast: config.Name += '_FastOptions'
config.InputFiles = os.path.abspath(args.inputrootfile)
# config.Tree = 'een_analyzer/{0}Tree'.format( particle.capitalize() )
config.Tree = 'een_analyzer/correction' # <-- May want to change this some time.
########################################
# BDT settings
########################################
if not args.fast:
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
"EventWeight=1",
]
else:
config.Options = [
"MinEvents=300", # Down from 200
"Shrinkage=0.2",
"NTrees=1000",
"MinSignificance=5.0", # Down from 5.0
"EventWeight=1",
]
# config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.Target = "(genEnergy * (trkMomentum*trkMomentum*trkMomentumRelError*trkMomentumRelError + (scRawEnergy+scPreshowerEnergy)*(scRawEnergy+scPreshowerEnergy)*resolution*resolution) / ( (scRawEnergy+scPreshowerEnergy)*response*trkMomentum*trkMomentum*trkMomentumRelError*trkMomentumRelError + trkMomentum*(scRawEnergy+scPreshowerEnergy)*(scRawEnergy+scPreshowerEnergy)*resolution*resolution ))"
# Probably neither of these are necessary
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.HistoConfig = "jobs/dummy_Histo.config"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
if region == 'EB':
config.DoEB = "True"
else:
config.DoEB = "False"
# # ======================================
# # Sample division - need a part for the ECAL-only training, and a part for the combination
# # 80% for the main BDT - divide the sample in divideNumber pieces, and use all but one piece for the main BDT
# divideNumber = 3
# config.CutBase = "eventNumber%{0}!=0".format( divideNumber )
# # 10% for combination, 10% for error
# config.CutComb = "eventNumber%{0}==0 && eventNumber%{1}==0".format( divideNumber, 2*divideNumber )
# config.CutError = "eventNumber%{0}==0 && eventNumber%{1}!=0".format( divideNumber, 2*divideNumber )
# config.CutBase += " && NtupID<5000"
# config.CutComb += " && NtupID<5000"
# config.CutError += " && NtupID<5000"
# Limit number of events in training
# config.CutBase = "NtupID<1000"
# config.CutComb = "NtupID<1000"
# config.CutError = "NtupID<1000"
# Pre-selected events have this variable set to 1.0
config.CutBase = "(ptWeightCut)"
config.CutComb = "(ptWeightCut)"
config.CutError = "(ptWeightCut)"
########################################
# Order tree branches
########################################
common_vars = [
"(scRawEnergy+scPreshowerEnergy)*response",
"resolution/response",
"trkMomentumRelError",
"trkMomentum/((scRawEnergy+scPreshowerEnergy)*response)",
"eleEcalDriven",
"fbrem",
"gsfchi2",
"gsfndof",
"trkEta",
"trkPhi",
]
config.VariablesEB = common_vars + [
]
config.VariablesEE = common_vars + [
]
if Verbose:
print '\n' + '-'*70
print 'Making config file ' + config.Name + '.config'
print ' Using the following branches for EE:'
print ' ' + '\n '.join( config.VariablesEE )
print ' Using the following branches for EB:'
print ' ' + '\n '.join( config.VariablesEB )
config.DoCombine = "False"
config.DoErrors = "False"
########################################
# Output
########################################
config.Parse()
return_configs.append( config )
return return_configs
def Make_conf(Verbose=True):
# Small testing samples -- do NOT use these for plots!
root_file = 'Ntup_Jun22_fullpt_testing_sample.root'
# root_file = 'Ntup_Jun22_lowpt_testing_sample.root'
# Low + high pt sample
# root_file = 'Ntup_Jun22_fullpt_training.root'
# Only low pt sample
# root_file = 'Ntup_Jun22_lowpt_training.root'
# ntup_path = os.path.join( '/data/userdata/rclsa/ElectronTrees/' )
ntup_path = os.path.join( os.environ['CMSSW_BASE'], 'src/NTuples' )
datestr = strftime( '%b%d' )
if not os.path.isdir( ntup_path ):
print 'Error: "{0}"" is not a directory'.format( ntup_path )
physical_path = lambda input_root_file: os.path.join( ntup_path, input_root_file )
return_configs = []
for particle in [ 'electron', 'photon' ]:
# Instantiate the Config class which prints a .config file
config = Config()
config.Name = 'Config_' + particle + '_' + datestr
config.InputFiles = physical_path( root_file )
config.Tree = 'een_analyzer/{0}Tree'.format( particle.capitalize() )
########################################
# BDT settings
########################################
config.Options = [
"MinEvents=200",
"Shrinkage=0.1",
"NTrees=1000",
"MinSignificance=5.0",
# "EventWeight=max( min(1,exp(-(genPt-50)/50)), 0.1 )", # <-- What to do?
"EventWeight=1", # <-- No one really likes the weights
]
config.Target = "genEnergy / ( scRawEnergy + scPreshowerEnergy )"
config.TargetError = "1.253*abs( BDTresponse - genEnergy / ( scRawEnergy + scPreshowerEnergy ) )"
config.HistoConfig = "jobs/dummy_Histo.config"
# config.CutBase = "eventNumber%2==0"
# Remove 2/7th of the events (reduces training from 70% to 50%)
config.CutBase = "eventNumber%2==0 && ( (eventNumber/2)%7==0 || (eventNumber/2)%7==1 )"
config.CutEB = "scIsEB"
config.CutEE = "!scIsEB"
config.CutError = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4==3)"
# Add an additional cut so that the regression is fast
# NtupIDcut = 10000
# config.CutBase += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutError += ' && (NtupID<={0})'.format( NtupIDcut )
# config.CutComb += ' && (NtupID<={0})'.format( NtupIDcut )
########################################
# Order tree branches
########################################
common_vars = [
# ======================================
# Common variables
# 'pt', # RCLSA: you cannot use the result of the previous training for the new one
# 'nVtx', # rho should be enough information for the BDT
'scRawEnergy',
# 'scEta', # Requires alignment information; use crystal number of the seed instead
# 'scPhi', # Requires alignment information; use crystal number of the seed instead
'scEtaWidth',
'scPhiWidth',
# 'scSeedRawEnergy/scRawEnergy', # RCLSA: Redundant with the one below
'full5x5_e5x5/scRawEnergy',
'hadronicOverEm',
'rhoValue',
'delEtaSeed',
'delPhiSeed',
# ======================================
# Showershape variables
# Use full 5x5 instead
# 'r9',
# 'eHorizontal',
# 'eVertical',
# 'sigmaIetaIeta',
# 'sigmaIetaIphi',
# 'sigmaIphiIphi',
# 'e5x5',
# 'e3x3',
# 'eMax',
# 'e2nd',
# 'eTop',
# 'eBottom',
# 'eLeft',
# 'eRight',
# 'e2x5Max',
# 'e2x5Left',
# 'e2x5Right',
# 'e2x5Top',
# 'e2x5Bottom',
# Normalization to scRawEnergy necessary?
'full5x5_r9',
# 'full5x5_eHorizontal', # RCLSA: Redundant
# 'full5x5_eVertical', # RCLSA: Redundant
'full5x5_sigmaIetaIeta',
'full5x5_sigmaIetaIphi',
'full5x5_sigmaIphiIphi',
# 'full5x5_e5x5', # RCLSA: Use ratios
# 'full5x5_e3x3/full5x5_e5x5', # RCLSA: Redundant, this is R9
'full5x5_eMax/full5x5_e5x5',
'full5x5_e2nd/full5x5_e5x5',
'full5x5_eTop/full5x5_e5x5',
'full5x5_eBottom/full5x5_e5x5',
'full5x5_eLeft/full5x5_e5x5',
'full5x5_eRight/full5x5_e5x5',
'full5x5_e2x5Max/full5x5_e5x5',
'full5x5_e2x5Left/full5x5_e5x5',
'full5x5_e2x5Right/full5x5_e5x5',
'full5x5_e2x5Top/full5x5_e5x5',
'full5x5_e2x5Bottom/full5x5_e5x5',
# ======================================
# Saturation variables
'N_SATURATEDXTALS',
# 'seedIsSaturated', # RCLSA: probably overkill
# 'seedCrystalEnergy/scSeedRawEnergy', # RCLSA: There is only 1/1e6 cases in which the max energy is not the seed
# ======================================
# Cluster variables
'N_ECALClusters',
# 'clusterMaxDR', # RCLSA Very mismodelled variables
# 'clusterMaxDRDPhi',
# 'clusterMaxDRDEta',
# 'clusterMaxDRRawEnergy',
'clusterRawEnergy[0]/scRawEnergy',
'clusterRawEnergy[1]/scRawEnergy',
'clusterRawEnergy[2]/scRawEnergy',
'clusterDPhiToSeed[0]',
'clusterDPhiToSeed[1]',
'clusterDPhiToSeed[2]',
'clusterDEtaToSeed[0]',
'clusterDEtaToSeed[1]',
'clusterDEtaToSeed[2]',
]
config.VariablesEB = common_vars + [
# 'cryEtaCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryPhiCoordinate', # Requires alignment information; use crystal number of the seed instead
'iEtaCoordinate',
'iPhiCoordinate',
'iEtaMod5',
'iPhiMod2',
'iEtaMod20',
'iPhiMod20',
]
config.VariablesEE = common_vars + [
# 'cryXCoordinate', # Requires alignment information; use crystal number of the seed instead
# 'cryYCoordinate', # Requires alignment information; use crystal number of the seed instead
'iXCoordinate',
'iYCoordinate',
'scPreshowerEnergy/scRawEnergy',
'preshowerEnergyPlane1/scRawEnergy',
'preshowerEnergyPlane2/scRawEnergy',
]
if Verbose:
print 'Using the following branches for EE:'
print ' ' + '\n '.join( config.VariablesEE )
print 'Using the following branches for EB:'
print ' ' + '\n '.join( config.VariablesEB )
########################################
# Ep combination
########################################
# Only do the combination for the electron
if particle == 'electron':
config.DoCombine = "True"
config.TargetComb = "( genEnergy - ( scRawEnergy + scPreshowerEnergy )*BDTresponse ) / ( trkMomentum - ( scRawEnergy + scPreshowerEnergy )*BDTresponse )"
config.CutComb = "(eventNumber%2!=0) && (((eventNumber-1)/2)%4!=3)"
config.VariablesComb = [
'( scRawEnergy + scPreshowerEnergy ) * BDTresponse',
'BDTerror/BDTresponse',
'trkMomentum',
'trkMomentumRelError',
'BDTerror/BDTresponse/trkMomentumRelError',
'( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum',
( '( scRawEnergy + scPreshowerEnergy )*BDTresponse/trkMomentum *' +
'sqrt( BDTerror/BDTresponse*BDTerror/BDTresponse + trkMomentumRelError*trkMomentumRelError)' ),
'eleEcalDriven',
'eleTrackerDriven',
'eleClass',
'scIsEB',
]
else:
config.DoCombine = "False"
########################################
# Output
########################################
if Verbose:
# Print all branches as a check
print "\nAll branches in root file:"
Read_branches_from_rootfile( physical_path(root_file) , config.Tree )
config.Parse()
# # Test if the config file can be read by ROOT TEnv
# print '\nReading in {0} and trying ROOT.TEnv( ..., 0 ):'.format( out_filename )
# I_TEnv = ROOT.TEnv()
# I_TEnv.ReadFile( out_filename, 0 )
# I_TEnv.Print()
# print 'Exited normally'
# print '='*70
# print
return_configs.append( config )
return return_configs
