Tau Tagging on a PFJet

Introduction

Tau lepton having a mass of 1.777 GeV, is the only lepton heavy enough to decay into hadrons. As depicted in the pi-chart, in about one third of the cases τ’s decay leptonically to a muon (τ_μ) or an electron (τ_e) with two neutrinos, and are reconstructed and identified with the usual techniques for muons and electrons. In the remaining cases, τ leptons decay hadronically, to a combination of charged and neutral mesons with a τ_ν. Hadronically decaying τ’s, denoted by τ_h, are reconstructed and identified with the hadrons-plus-strips (HPS) algorithm, which was developed for use in the LHC Run-1. The key challenge that this algorithm has to face is the distinction between genuine τ_h, and quark and gluon jets, which are copiously produced in QCD multijet process and can be misidentified as τ_h. The main handle for reducing these jet→τ_h misidentification backgrounds is to utilize the fact that the particles produced in τ_h decays are of lower multiplicity, deposit energy in a narrow region compared to a quark or gluon jet, and are typically isolated with respect to other particles in the event. In some physics analyses, the misidentification of electrons or muons as τ_h candidates may constitute a sizeable background as well. Therefore, HPS algorithm has got various discriminators like isolation, against electrons and muons etc. to identify genuine hadronically decaying taus.

Tau Reconstruction and Identification Software

The hadronic tau-jet candidates are created from ParticleFlow jets using hadron-plus-strip (HPS) algorithm. The tau reconstruction code can be found in:

• DataFormats/TauReco - Definition of the dataformats (see event content).
• RecoTauTag/RecoTau - Implementation of the discriminators and decay mode finder.

The details on HPS algorithm can be found at:

• J. Instrum. 7 (2012) P01001 Tau 2010 Commissioning paper.
• J. Instrum. 11 (2016) P01019 LHC Run 1 (at 8 TeV) Tau paper.

Creation

Rerun the tau sequences, load the tau steering sequence in your cfg file: process.load("RecoTauTag.Configuration.RecoPFTauTag_cff")

On RECO level reco::PFTaus are used to store all possible taus jets. The result of the various discriminators is stored in reco::PFTauDiscriminator.

The initial PFTaus collection will first contain one tau candidate for each reconstructed PFJet (using the anti-kt algorithm with a cone size of 0.5). The jet direction is set to the direction of the leading charged hadron in the jet (highest pT), which should be within ∆R = 0.1 with respect to the jet axis. In this stage the signal- and isolation cones as well as the decay mode are defined. Subsequently the HPS algorithm create PFTauDiscriminators which can be used to select a collection of PFTaus.

Explanation of signal- and isolation cones Hide

Usage

in your event loop (e.g. the analyze(const edm::Event& iEvent) method of your analyzer) do

Recipe Hide

// read PFTau collection (set pfTauToken_ in to an edm::InputTag before)

     edm::EDGetTokenT<reco::PFTauCollection> pfTauToken_;
     edm::Handle<reco::PFTauCollection> pfTaus;
     iEvent.getByToken(pfTauToken_,pfTaus);

//read one PFTauDiscriminator (set discriminatorToken_ in to an edm::InputTag before)

    edm::EDGetTokenT<reco::PFTauDiscriminator> discriminatorToken_;
    edm::Handle<reco::PFTauDiscriminator> discriminator;
    iEvent.getByToken(discriminatorToken_,discriminator);

// loop over taus
    for ( unsigned iTau = 0; iTau < pfTaus->size(); ++iTau ) {
        reco::PFTauRef tauCandidate(pfTaus, iTau);
// check if tau candidate has passed discriminator
        if( (*discriminator)[tauCandidate] > 0.5 ){
        // do something with your candidate
        }
    }

Discriminators

The PFTauDiscriminators are used store the result of the various tau tagging algorithms and select jets that likely originate from a hadronic tau decay. By definition they are real numbers between 0 and 1 where higher numbers indicate a more positive outcome of the discriminator. Note that most discriminators are in fact binary thus taking just values of 0 (=did not pass) and 1 (=passed).

The names PFTauDiscriminator collections follow the <algorithm-prefix> + <discriminator name> convention. (e.g. the AgainstElectron collection can be accessed via hpsPFTauDiscriminationAgainstElectron)

Legacy Tau ID (Run I)

The <algorithm-prefix> is hpsPFTauDiscrimination

Name	binary?	Description
ByLooseElectronRejection	yes	electron pion MVA discriminator < 0.6
ByMediumElectronRejection	yes	electron pion MVA discriminator < -0.1 and not 1.4442 < η < 1.566 (both positive & negative eta values)
ByTightElectronRejection	yes	electron pion MVA discriminator < -0.1 and not 1.4442 < η < 1.566 (both positive & negative eta values) and Brem pattern cuts (see AN-10-387)
ByMVA3Loose/Medium/Tight/VTightElectronRejection	yes	anti-electron MVA discriminator with improved training (see talk)
ByMVA3VTightElectronRejection	yes	anti-electron MVA discriminator with improved training (same efficiency as "HCP 2012 working point" (see talk)
ByLooseMuonRejection	yes	Tau Lead Track not matched to chamber hits
ByMediumMuonRejection	yes	Tau Lead Track not matched to global/tracker muon
ByTightMuonRejection	yes	Tau Lead Track not matched to global/tracker muon and large enough energy deposit in ECAL + HCAL
ByLooseMuonRejection2	yes	Same as AgainstMuonLoose
ByMediumMuonRejection2	yes	Loose2 && no DT, CSC or RPC Hits in last 2 Stations
ByTightMuonRejection2	yes	Medium2 && large enough energy deposit in ECAL + HCAL in 1 prong + 0 strip decay mode (Σ(ECAL+HCAL) > 0.2 * pT)
ByDecayModeFinding	yes	You will always want to use this (see AN-10-82 )
ByVLooseIsolation	yes	isolation cone of 0.3 , no PF Charged Candidates with pT > 1.5 GeV/c and no PF Gamma candidates with ET > 2.0 GeV
ByVLooseCombinedIsolationDBSumPtCorr	yes	isolation cone of 0.3 , Delta Beta corrected sum pT of PF charged and PF gamma isolation candidates (pT > 0.5 GeV) less than 3 GeV
ByLooseCombinedIsolationDBSumPtCorr	yes	isolation cone of 0.5 , Delta Beta corrected sum pT of PF charged and PF gamma isolation candidates (pT > 0.5 GeV) less than 2 GeV
ByMediumCombinedIsolationDBSumPtCorr	yes	isolation cone of 0.5 , Delta Beta corrected sum pT of PF charged and PF gamma isolation candidates (pT > 0.5 GeV) less than 1 GeV
ByTightCombinedIsolationDBSumPtCorr	yes	isolation cone of 0.5 , Delta Beta corrected sum pT of PF charged and PF gamma isolation candidates (pT > 0.5 GeV) less than 0.8 GeV
ByLooseCombinedIsolationDBSumPtCorr3Hits	yes	same as ByLooseCombinedIsolationDBSumPtCorr but requiring 3 hits (instead of 8) on track of isolation candidates
ByMediumCombinedIsolationDBSumPtCorr3Hits	yes	same as ByMediumCombinedIsolationDBSumPtCorr but requiring 3 hits (instead of 8) on track of isolation candidates
ByTightCombinedIsolationDBSumPtCorr3Hits	yes	same as ByTightCombinedIsolationDBSumPtCorr but requiring 3 hits (instead of 8) on track of isolation candidates
ByLooseIsolationMVA	yes	BDT based selection using isolation in rings around tau direction and shower shape variables
ByMediumIsolationMVA	yes	BDT based selection using isolation in rings around tau direction and shower shape variables
ByTightIsolationMVA	yes	BDT based selection using isolation in rings around tau direction and shower shape variables
ByIsolationMVAraw	no	output of BDT based selection using isolation in rings around tau direction and shower shape variables
ByLooseIsolationMVA2	yes	same as ByLooseIsolationMVA with new training and improved performance
ByMediumIsolationMVA2	yes	same as ByMediumIsolationMVA with new training and improved performance
ByTightIsolationMVA2	yes	same as ByTightIsolationMVA with new training and improved performance
ByIsolationMVA2raw	no	output of "MVA2" BDT discriminator

Isolation discriminators usage

Loosening of quality cuts on tracks of isolation candidates means that more of them can enter isolation pt. This results in slightly decreased efficiency of tau identification but also considerably decreased jet fake rate. More details in this talk.

The combined isolation is recommended for all taus, while MVA based isolation can provide better performance for low pt (< 100 GeV) taus. Since RecoTauTag/RecoTau V01-04-23 (V01-04-23-4XX-00 for 4XX analysis) a new training for MVA isolation is available. The discriminators have "IsolationMVA2" in their name.

"MVA3" anti-electron discriminators

The new MVA training provides working points with decreased electron fake rate when keeping the same efficiency as for previous training. All MVA3 discriminators veto tau candidates in a crack region between barrel and endcap, so they are NOT RECOMMENDED if you are using tau id for TAU VETO in your analysis.

"AgainstMuon2" discriminators

A drop of efficiency of anti-muon discriminator have been observed in high p_T. The fix is available in "Muon2" discriminators. More details in ( talk 1 and talk 2)

HPS decay mode definition Hide

int A = tau.signalPFChargedHadrCands().size()
int B = tau.signalPFGammaCands().size()

Tau ID 2014 (preparation for Run II)

The <algorithm-prefix> is hpsPFTauDiscrimination

Name	binary?	Description
ByLooseElectronRejection	yes	electron pion MVA discriminator < 0.6
ByMediumElectronRejection	yes	electron pion MVA discriminator < -0.1 and not 1.4442 < η < 1.566 (both positive & negative eta values)
ByTightElectronRejection	yes	electron pion MVA discriminator < -0.1 and not 1.4442 < η < 1.566 (both positive & negative eta values) and Brem pattern cuts (see AN-10-387)
ByMVA5(Loose/Medium/Tight/VTight)ElectronRejection	yes	anti-electron MVA discriminator with new training
ByLooseMuonRejection	yes	Tau Lead Track not matched to chamber hits
ByMediumMuonRejection	yes	Tau Lead Track not matched to global/tracker muon
ByTightMuonRejection	yes	Tau Lead Track not matched to global/tracker muon and energy deposit in ECAL + HCAL exceeding 0.2 times Lead Track momentum
ByLooseMuonRejection3	yes	Tau Lead Track not matched to more than one segment in muon system, energy deposit in ECAL + HCAL at least 0.2 times Lead Track momentum
ByTightMuonRejection3	yes	Tau Lead Track not matched to more than one segment or hits in the outermost two stations of the muon system, energy deposit in ECAL + HCAL at least 0.2 times Lead Track momentum
ByMVA(Loose/Medium/Tight)MuonRejection	yes	BDT based anti-muon discriminator
ByMVArawMuonRejection	no	raw MVA output of BDT based anti-muon discriminator
ByDecayModeFinding	yes	You will always want to use this (see AN-10-82 )
ByVLooseIsolation	yes	isolation cone of 0.3 , no PF Charged Candidates with pT > 1.5 GeV/c and no PF Gamma candidates with ET > 2.0 GeV
ByVLooseCombinedIsolationDBSumPtCorr	yes	isolation cone of 0.3 , Delta Beta corrected sum pT of PF charged and PF gamma isolation candidates (pT > 0.5 GeV) less than 3 GeV
ByLooseCombinedIsolationDBSumPtCorr	yes	isolation cone of 0.5 , Delta Beta corrected sum pT of PF charged and PF gamma isolation candidates (pT > 0.5 GeV) less than 2 GeV
ByMediumCombinedIsolationDBSumPtCorr	yes	isolation cone of 0.5 , Delta Beta corrected sum pT of PF charged and PF gamma isolation candidates (pT > 0.5 GeV) less than 1 GeV
ByTightCombinedIsolationDBSumPtCorr	yes	isolation cone of 0.5 , Delta Beta corrected sum pT of PF charged and PF gamma isolation candidates (pT > 0.5 GeV) less than 0.8 GeV
ByLooseCombinedIsolationDBSumPtCorr3Hits	yes	same as ByLooseCombinedIsolationDBSumPtCorr but requiring 3 hits (instead of 8) on track of isolation candidates
ByMediumCombinedIsolationDBSumPtCorr3Hits	yes	same as ByMediumCombinedIsolationDBSumPtCorr but requiring 3 hits (instead of 8) on track of isolation candidates
ByTightCombinedIsolationDBSumPtCorr3Hits	yes	same as ByTightCombinedIsolationDBSumPtCorr but requiring 3 hits (instead of 8) on track of isolation candidates
By(VLoose/Loose/Medium/Tight/VTight/VVTight)IsolationMVA3oldDMwoLT	yes	BDT based tau ID discriminator based on isolation Pt sums, trained on 1-prong and 3-prong tau candidates
ByIsolationMVA3oldDMwoLTraw	no	raw MVA output of BDT based tau ID discriminator based on isolation Pt sums, trained on 1-prong and 3-prong tau candidates
By(VLoose/Loose/Medium/Tight/VTight/VVTight)IsolationMVA3oldDMwLT	yes	BDT based tau ID discriminator based on isolation Pt sums plus tau lifetime information, trained on 1-prong and 3-prong tau candidates
ByIsolationMVA3oldDMwLTraw	no	raw MVA output of BDT based tau ID discriminator based on isolation Pt sums plus tau lifetime information, trained on 1-prong and 3-prong tau candidates
By(VLoose/Loose/Medium/Tight/VTight/VVTight)IsolationMVA3newDMwoLT	yes	BDT based tau ID discriminator based on isolation Pt sums, trained on 1-prong, "2-prong" and 3-prong tau candidates
ByIsolationMVA3newDMwoLTraw	no	raw MVA output of BDT based tau ID discriminator based on isolation Pt sums, trained on 1-prong, "2-prong" and 3-prong tau candidates
By(VLoose/Loose/Medium/Tight/VTight/VVTight)IsolationMVA3newDMwLT	yes	BDT based tau ID discriminator based on isolation Pt sums plus tau lifetime information, trained on 1-prong, "2-prong" and 3-prong tau candidates
ByIsolationMVA3newDMwLTraw	no	raw MVA output of BDT based tau ID discriminator based on isolation Pt sums plus tau lifetime information, trained on 1-prong, "2-prong" and 3-prong tau candidates

How to get latest TauID on MiniAOD event content

While it is not possible to fully rebuild taus from jets given MiniAOD event content, it is possible to recompute the BDT output of both isolation and anti-electron discriminators for new trainings made available starting from CMSSW 8_0_X onwards. While for releases from 8_1_X onwards the necessary infrastructure is included in official CMSSW releases, for 8_0_X one has to merge developments from the following cms-tau-pog branch: cms-tau-pog/CMSSW_8_0_X_tau-pog_miniAOD-backport-tauID . In your CMSSW_8_0_X/src/ area do:

git cms-merge-topic -u cms-tau-pog:CMSSW_8_0_X_tau-pog_miniAOD-backport-tauID

The -u is necessary to avoid that git checks out all packages that depend on any of the ones touched in this branch since this would lead to a very long compilation. Then compile everything.

In order to be able to access the latest and greatest BDT output for the isolation discriminators and save it in your ntuples, you need to add a sequence to your python config file and some code to your analyzer itself. You can find an example analyzer and the corrsponding config file in RecoTauTag/RecoTau/test . Below, a code example for including the new training with old decay modes is shown. The procedure is the same for the training with new decay modes. Please refer to the TauIDRecommendation13TeV TWiki for the lines that need to be changed.

Additions to python config file for new MVA isolation Hide

from RecoTauTag.RecoTau.TauDiscriminatorTools import noPrediscriminants
process.load('RecoTauTag.Configuration.loadRecoTauTagMVAsFromPrepDB_cfi')
from RecoTauTag.RecoTau.PATTauDiscriminationByMVAIsolationRun2_cff import *

process.rerunDiscriminationByIsolationMVArun2v1raw = patDiscriminationByIsolationMVArun2v1raw.clone(
   PATTauProducer = cms.InputTag('slimmedTaus'),
   Prediscriminants = noPrediscriminants,
   loadMVAfromDB = cms.bool(True),
   mvaName = cms.string("RecoTauTag_tauIdMVAIsoDBoldDMwLT2016v1"), # name of the training you want to use
   mvaOpt = cms.string("DBoldDMwLT"), # option you want to use for your training (i.e., which variables are used to compute the BDT score)
   requireDecayMode = cms.bool(True),
   verbosity = cms.int32(0)
)

process.rerunDiscriminationByIsolationMVArun2v1VLoose = patDiscriminationByIsolationMVArun2v1VLoose.clone(
   PATTauProducer = cms.InputTag('slimmedTaus'),    
   Prediscriminants = noPrediscriminants,
   toMultiplex = cms.InputTag('rerunDiscriminationByIsolationMVArun2v1raw'),
   key = cms.InputTag('rerunDiscriminationByIsolationMVArun2v1raw:category'),
   loadMVAfromDB = cms.bool(True),
   mvaOutput_normalization = cms.string("RecoTauTag_tauIdMVAIsoDBoldDMwLT2016v1_mvaOutput_normalization"), # normalization fo the training you want to use
   mapping = cms.VPSet(
      cms.PSet(
         category = cms.uint32(0),
         cut = cms.string("RecoTauTag_tauIdMVAIsoDBoldDMwLT2016v1_WPEff90"), # this is the name of the working point you want to use
         variable = cms.string("pt"),
      )
   )
)

# here we produce all the other working points for the training
process.rerunDiscriminationByIsolationMVArun2v1Loose = process.rerunDiscriminationByIsolationMVArun2v1VLoose.clone()
process.rerunDiscriminationByIsolationMVArun2v1Loose.mapping[0].cut = cms.string("RecoTauTag_tauIdMVAIsoDBoldDMwLT2016v1_WPEff80")
process.rerunDiscriminationByIsolationMVArun2v1Medium = process.rerunDiscriminationByIsolationMVArun2v1VLoose.clone()
process.rerunDiscriminationByIsolationMVArun2v1Medium.mapping[0].cut = cms.string("RecoTauTag_tauIdMVAIsoDBoldDMwLT2016v1_WPEff70")
process.rerunDiscriminationByIsolationMVArun2v1Tight = process.rerunDiscriminationByIsolationMVArun2v1VLoose.clone()
process.rerunDiscriminationByIsolationMVArun2v1Tight.mapping[0].cut = cms.string("RecoTauTag_tauIdMVAIsoDBoldDMwLT2016v1_WPEff60")
process.rerunDiscriminationByIsolationMVArun2v1VTight = process.rerunDiscriminationByIsolationMVArun2v1VLoose.clone()
process.rerunDiscriminationByIsolationMVArun2v1VTight.mapping[0].cut = cms.string("RecoTauTag_tauIdMVAIsoDBoldDMwLT2016v1_WPEff50")
process.rerunDiscriminationByIsolationMVArun2v1VVTight = process.rerunDiscriminationByIsolationMVArun2v1VLoose.clone()
process.rerunDiscriminationByIsolationMVArun2v1VVTight.mapping[0].cut = cms.string("RecoTauTag_tauIdMVAIsoDBoldDMwLT2016v1_WPEff40")

# this sequence has to be included in your cms.Path() before your analyzer which accesses the new variables is called.
process.rerunMvaIsolation2SeqRun2 = cms.Sequence(
   process.rerunDiscriminationByIsolationMVArun2v1raw
   *process.rerunDiscriminationByIsolationMVArun2v1VLoose
   *process.rerunDiscriminationByIsolationMVArun2v1Loose
   *process.rerunDiscriminationByIsolationMVArun2v1Medium
   *process.rerunDiscriminationByIsolationMVArun2v1Tight
   *process.rerunDiscriminationByIsolationMVArun2v1VTight
   *process.rerunDiscriminationByIsolationMVArun2v1VVTight
)

# embed new id's into new tau collection
embedID = cms.EDProducer("PATTauIDEmbedder",
   src = cms.InputTag('slimmedTaus'),
   tauIDSources = cms.PSet(
      byIsolationMVArun2v1DBoldDMwLTrawNew = cms.InputTag('rerunDiscriminationByIsolationMVArun2v1raw'),
      byVLooseIsolationMVArun2v1DBoldDMwLTNew = cms.InputTag('rerunDiscriminationByIsolationMVArun2v1VLoose'),
      byLooseIsolationMVArun2v1DBoldDMwLTNew = cms.InputTag('rerunDiscriminationByIsolationMVArun2v1Loose'),
      byMediumIsolationMVArun2v1DBoldDMwLTNew = cms.InputTag('rerunDiscriminationByIsolationMVArun2v1Medium'),
      byTightIsolationMVArun2v1DBoldDMwLTNew = cms.InputTag('rerunDiscriminationByIsolationMVArun2v1Tight'),
      byVTightIsolationMVArun2v1DBoldDMwLTNew = cms.InputTag('rerunDiscriminationByIsolationMVArun2v1VTight'),
      byVVTightIsolationMVArun2v1DBoldDMwLTNew = cms.InputTag('rerunDiscriminationByIsolationMVArun2v1VVTight'),
      . . . (other discriminators like anti-electron),
      ),
   )
setattr(process, "NewTauIDsEmbedded", embedID)

process.p = cms.Path(
   . . . (other processes)
   *process.rerunMvaIsolation2SeqRun2
   *getattr(process, "NewTauIDsEmbedded")
   . . . (for example you ntuple creation process)
)

The python configuration to be added to include new trainings of the anti-electron discriminator is shown below.

Additions to python config file for new anti-electron MVA Hide

process.load('RecoTauTag.Configuration.loadRecoTauTagMVAsFromPrepDB_cfi')
from RecoTauTag.RecoTau.PATTauDiscriminationAgainstElectronMVA6_cfi import *

process.rerunDiscriminationAgainstElectronMVA6 = patTauDiscriminationAgainstElectronMVA6.clone(
   PATTauProducer = cms.InputTag('slimmedTaus'),
   Prediscriminants = noPrediscriminants,
   #Prediscriminants = requireLeadTrack,
   loadMVAfromDB = cms.bool(True),
   returnMVA = cms.bool(True),
   method = cms.string("BDTG"),
   mvaName_NoEleMatch_woGwoGSF_BL = cms.string("RecoTauTag_antiElectronMVA6v1_gbr_NoEleMatch_woGwoGSF_BL"),
   mvaName_NoEleMatch_wGwoGSF_BL = cms.string("RecoTauTag_antiElectronMVA6v1_gbr_NoEleMatch_wGwoGSF_BL"),
   mvaName_woGwGSF_BL = cms.string("RecoTauTag_antiElectronMVA6v1_gbr_woGwGSF_BL"),
   mvaName_wGwGSF_BL = cms.string("RecoTauTag_antiElectronMVA6v1_gbr_wGwGSF_BL"),
   mvaName_NoEleMatch_woGwoGSF_EC = cms.string("RecoTauTag_antiElectronMVA6v1_gbr_NoEleMatch_woGwoGSF_EC"),
   mvaName_NoEleMatch_wGwoGSF_EC = cms.string("RecoTauTag_antiElectronMVA6v1_gbr_NoEleMatch_wGwoGSF_EC"),
   mvaName_woGwGSF_EC = cms.string("RecoTauTag_antiElectronMVA6v1_gbr_woGwGSF_EC"),
   mvaName_wGwGSF_EC = cms.string("RecoTauTag_antiElectronMVA6v1_gbr_wGwGSF_EC"),
   minMVANoEleMatchWOgWOgsfBL = cms.double(0.0),
   minMVANoEleMatchWgWOgsfBL  = cms.double(0.0),
   minMVAWOgWgsfBL            = cms.double(0.0),
   minMVAWgWgsfBL             = cms.double(0.0),
   minMVANoEleMatchWOgWOgsfEC = cms.double(0.0),
   minMVANoEleMatchWgWOgsfEC  = cms.double(0.0),
   minMVAWOgWgsfEC            = cms.double(0.0),
   minMVAWgWgsfEC             = cms.double(0.0),
   srcElectrons = cms.InputTag('slimmedElectrons')
)
# embed new id's into new tau collection
embedID = cms.EDProducer("PATTauIDEmbedder",
   src = cms.InputTag('slimmedTaus'),
   tauIDSources = cms.PSet(
      . . . (other new discriminators like isolation),
      againstElectronMVA6RawNew = cms.InputTag('rerunDiscriminationAgainstElectronMVA6')
      ),
   )
setattr(process, "NewTauIDsEmbedded", embedID)

process.p = cms.Path(
   . . . (other processes)
   *process.rerunDiscriminationAgainstElectronMVA6
   *getattr(process, "NewTauIDsEmbedded")
   . . . (for example you ntuple creation process)
)

Please be mindful that if you want to include new isolation and anti-electron discriminators at the same time, things like the PATTauIDEmbedder need only be run once. Just reorder/change the example python configuration snippets accordingly.

Once the new discriminators are embedded into the new tau collection (in this example called "NewTauIDsEmbedded") one can simply retrieve them via the pat::Tau::tauID function in a loop over the collection like so:

float newIsolationMVArawValue = tau->tauID("byIsolationMVArun2v1DBoldDMwLTrawNew");
float newAntiElectronMVArawValue = tau->tauID("againstElectronMVA6RawNew");

Rerunning of the TauID on AOD event content

The tau collections contained in the simulated samples are often outdated. The most recent tau ID is in general contained in the releases, so to benefit from all new features and bugfixes, you should re-run the PFTau sequence on RECO/AOD. Because the the most up-to-date software is almost always contained in the production releases, there is no need to merge in any code from other repositories.

To re-run the tau sequence in, you need to add following few lines to your config file

process.load("RecoTauTag.Configuration.RecoPFTauTag_cff") #loading the configuration

from PhysicsTools.PatAlgos.tools.tauTools import * # to get the most up-to-date discriminators when using patTaus
switchToPFTauHPS(process) # this line and the one above can be ignored when running on reco::PFTauCollection directly
...
process.p = cms.Path(   #adding to the path
....
        process.PFTau*
....
)

When running in un-scheduled mode it is enough to add process.load("RecoTauTag.Configuration.RecoPFTauTag_cff") to the config file.

Important links for detailed information

More detailed information can be found at: twiki 1 and twiki 2.

Review status

Responsible: NitishDhingra
-- NitishDhingra - 2017-09-07