General Information

B-tagging code

ttH references

Tracking/vertexing code and references

MCP and EGamma code

Presentations - recent

Presentations - older

Documentation

References

ttbar sample xAOD dsid

mc16_13TeV:mc16_13TeV.410470.PhPy8EG_A14_ttbar_hdamp258p75_nonallhad.merge.AOD.e6337_e5984_s3126_r9364_r9315 mc16_13TeV:mc16_13TeV.410470.PhPy8EG_A14_ttbar_hdamp258p75_nonallhad.merge.AOD.e6337_e5984_s3126_r10201_r10210 mc16_13TeV:mc16_13TeV.410470.PhPy8EG_A14_ttbar_hdamp258p75_nonallhad.merge.AOD.e6337_e5984_s3126_r10724_r10726

ttbar sample MUON5 DxAOD dsid

mc16_13TeV.410470.PhPy8EG_A14_ttbar_hdamp258p75_nonallhad.deriv.DAOD_MUON5.e6337_e5984_s3126_r9364_r9315_p3980 22.411 TB, 119432000 mc16_13TeV.410470.PhPy8EG_A14_ttbar_hdamp258p75_nonallhad.deriv.DAOD_MUON5.e6337_e5984_s3126_r10201_r10210_p3980 mc16_13TeV.410470.PhPy8EG_A14_ttbar_hdamp258p75_nonallhad.deriv.DAOD_MUON5.e6337_e5984_s3126_r10201_r10210_p3980

Zmumugam sample xAOD dsid

mc16_13TeV:mc16_13TeV.366145.Sh_224_NN30NNLO_mumugamma_LO_pty_7_15.merge.AOD.e7006_e5984_s3126_r9364_r9315 Total events : 999000 mc16_13TeV:mc16_13TeV.366146.Sh_224_NN30NNLO_mumugamma_LO_pty_15_35.merge.AOD.e7006_e5984_s3126_r9364_r9315 Total events : 3995000 mc16_13TeV:mc16_13TeV.366147.Sh_224_NN30NNLO_mumugamma_LO_pty_35_70.merge.AOD.e7006_e5984_s3126_r9364_r9315 Total events : 499000 mc16_13TeV:mc16_13TeV.366148.Sh_224_NN30NNLO_mumugamma_LO_pty_70_140.merge.AOD.e7006_e5984_s3126_r9364_r9315 Total events : 250000 mc16_13TeV:mc16_13TeV.366149.Sh_224_NN30NNLO_mumugamma_LO_pty_140_E_CMS.merge.AOD.e7006_e5984_s3126_r9364_r9315 Total events : 250000 mc16_13TeV:mc16_13TeV.366145.Sh_224_NN30NNLO_mumugamma_LO_pty_7_15.merge.AOD.e7006_e5984_s3126_r10201_r10210 Total events : 1246000 mc16_13TeV:mc16_13TeV.366146.Sh_224_NN30NNLO_mumugamma_LO_pty_15_35.merge.AOD.e7006_e5984_s3126_r10201_r10210 Total events : 4985000 mc16_13TeV:mc16_13TeV.366147.Sh_224_NN30NNLO_mumugamma_LO_pty_35_70.merge.AOD.e7006_e5984_s3126_r10201_r10210 Total events : 624000 mc16_13TeV:mc16_13TeV.366148.Sh_224_NN30NNLO_mumugamma_LO_pty_70_140.merge.AOD.e7006_e5984_s3126_r10201_r10210 Total events : 319000 mc16_13TeV:mc16_13TeV.366149.Sh_224_NN30NNLO_mumugamma_LO_pty_140_E_CMS.merge.AOD.e7006_e5984_s3126_r10201_r10210 Total events : 320000 mc16_13TeV:mc16_13TeV.366145.Sh_224_NN30NNLO_mumugamma_LO_pty_7_15.merge.AOD.e7006_e5984_s3126_r10724_r10726 Total events : 1670000 mc16_13TeV:mc16_13TeV.366146.Sh_224_NN30NNLO_mumugamma_LO_pty_15_35.merge.AOD.e7006_e5984_s3126_r10724_r10726 Total events : 6461000 mc16_13TeV:mc16_13TeV.366147.Sh_224_NN30NNLO_mumugamma_LO_pty_35_70.merge.AOD.e7006_e5984_s3126_r10724_r10726 Total events : 834000 mc16_13TeV:mc16_13TeV.366148.Sh_224_NN30NNLO_mumugamma_LO_pty_70_140.merge.AOD.e7006_e5984_s3126_r10724_r10726 Total events : 418000 mc16_13TeV:mc16_13TeV.366149.Sh_224_NN30NNLO_mumugamma_LO_pty_140_E_CMS.merge.AOD.e7006_e5984_s3126_r10724_r10726 Total events : 250000

RNN INPUT samples

  • Zgam mini-ntup samples (Full run-2) atint:/net/ustc_03/prompt/MININTUP/zgam_mc_fullrun2.root
  • Zgam mini-ntup samples atint:/net/ustc_03/prompt/MININTUP/zgam_mca[d,e]
  • Zgam ntup samples atint:/net/ustc_03/prompt/NTUP/zgam_mca[d,e]

Data

  • Penn EOS storage space:
ssh -XY pennww@lxplus.cern.ch
/eos/atlas/atlascerngroupdisk/penn-ww/

Setup atlas environment in release 21

acmSetup --sourcedir=../source AthDerivation,21.2,21.2.3.0
  • Checkout packages:
acm add_pkg athena/PhysicsAnalysis/DerivationFramework/DerivationFrameworkHiggs
acm add_pkg athena/PhysicsAnalysis/JetTagging/JetTagNonPromptLepton
  • Download AOD for testing (e.g. mc16_13TeV.410501.PowhegPythia8EvtGen_A14_ttbar_hdamp258p75_nonallhad.merge.AOD.e5458_s3126_r9364_r9315) and run derivation with command:
Reco_tf.py --inputAODFile input_AOD.pool.root --outputDAODFile output.pool.root --reductionConf HIGG8D1 --maxEvents 100
athena share/JetTagNonPromptLepton_decorate.py -c 'inputDir="{MUON5_file}";EvtMax=10'
  • When making merge requests be sure to add a "sweep:ignore" label, which stops git trying to merge the 21.2 development branch (the only place JetTagNonPromptLepton lives) with master. Also add "Derivation" label.

Summary of taggers

  • SV1
    • Uses loosest track selection with pT>400 MeV
  • JetFitter is probably most complex tagger
    • Uses tracks with pT > 786 MeV
    • Difficult to change configuration and to understand and interpret results
    • Requires substantial time investment to learn how to operate
  • IP3D is most optimal for low and medium pT jets (based on BDT weights of input variables)
    • Uses tracks with pT > 1000 MeV
    • Requires training files with PDFs
    • Produce training files by setting doComputeReference=True
    • Can we train only for our specific topology?
    • Important information in sign of impact parameter
  • Variable size cone for track selection
    • 0.45 for jet pT=20 GeV

Initial plan

  • Make full b-tagging ntuples with all tracks
    • ReduceInfo =False
    • Saves all reconstructed tracks
    • Flag for which tagger used this track
    • Flag for whether track matches a particle from b-hadron
  • Study which tracks are used by IPD3
    • Fraction of b-hadron tracks used for different selection
  • Change track selection for IP3D
    • Create new references
    • Rerun with new references

Proposed tasks for b-tagging algorithm group

  • Compare MV2 with standalone IP3D
    • Can IP3D alone perform as well as MV2 for this topology?
    • Add default IP3D to our ntuples
    • Also compare returned IP3D with MV2
  • Switch to using b-tagging performance ntuples with all tracks
    • Check that muon variables are saved: pT, eta, phi, isolation and impact variables
    • Add muon truth matching and parent truth information
  • Study track selection for IP3D
  • Train new IP3D with updated track selection
    • Create new references and repeat our study
  • Longer term issues
    • Does fake rejection depends on pileup? What is pileup and underlying event systematic?
    • Study electrons

Ideas for improvements

  • Can we reconstruct K0 to pi+pi- decays? Does this help to identify B decays?
  • Does it help to select a reconstructed ID track which has a highest impact parameter when combined with muon track?
  • Make event displays to understand how b-tag veto can be improved
  • Systematic uncertainty: correlations between b-tagging weight and isolation variables


Tasks

Current task list

  • Develop and document standalone tool - if necesseary
    • Code and instructions for adding necessary input variables to derivations: track jets and flavour tagging
    • Code and instructions for standalone dual tool that runs on xAOD
  • Optimise saving of input and output variables in xAOD
    • Different prefixes for input and output variables
    • Add python confirguration for saving only BDT weights into DxAOD (to save space for CP derivations)
  • Merge package branch and trunk

Working with data

  • The calculation of the fake efficiency of whatever algorithm we use can be done by
    • Looking at the inclusive muon spectrum, whilst subtracting the W and Z components to look at the multijet background.
    • Looking at the 2lSS ttbar control region: 2 or 3 jets, 2 same sign leptons. The problem with this is that there is a sizeable ttbarW background to disentangle - which would be done with Monte Carlo.
  • The easier of the two is looking at the control region - this is the method that we will use for ICHEP.
  • The other method is less biased but harder - to keep in mind for the long term.


To-do list (updated 12-01-17)

Truth study tasks

  • Do full truth study of lepton origins, including non-prompt and charge misid - done.
  • Plot BDT for electrons with direct tau parent to see if the BDT biases against taus which have similar lifetime to B hadrons - done.

BDT optimisation tasks

  • Add pT(lepton)/pT(jet) and dR(lepton, jet) variables in BDT to see how rejection improves - done
  • Have an "all-in-one" BDT including isolation, impact parameter etc. and a "b-tagging only" BDT - done

Ideas for updates

  • Check pTRel variable - like pT(lepton)/pT(jet)
  • Remove one of ip2 or ip3
  • Use both 410000 and 410009 for training to improve statistics
Edit | Attach | Watch | Print version | History: r98 < r97 < r96 < r95 < r94 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r98 - 2019-11-08 - RustemOspanov
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

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