News

HQT Software Task Force

Mandate and current activities

  • Facilitate communication among analysis regarding analysis software
  • Help with basic validation of software frameworks used in HQT
  • Perform basic cross-framework validation to ensure reproducibility of physics results from different frameworks

Contacts

Major analysis frameworks used in HQT

According to the early survey, frameworks used in individual analysis for Rel.20.7 and their corresponding derivations are listed in the following:

Analysis Framework Derivation(s) Processing time
tt resonance (1-lep.) AT-based EXOT4 ~1 WK
tt resonance (0-lep.) AT-based EXOT7+EXOT4+TOPQ1 ~1 WK
tb resonance (0-lep.) AT-based EXOT7 -
VLQ OS-ML AT-based EXOT4 ~2-4 WK
VLQ all-had. AT-based EXOT7 -
VLQ Zt+X high-MET ST-based SUSY5 -
VLQ Ht+X ST-based TOPQ1+TOPQ4 -
VLQ B->b/Zb CxAOD - in Development

Cross-Framework Comparison

Setups

Softwares

Validation Tool: https://gitlab.cern.ch/CrossFrameworkValidation/CrossFrameworkValidation

Frameworks

  Validated/Latest Version Contact Source Code Build Converge
AnalysisTop 21.2.19/21.2.37 hn-atlas-top-reconstruction@cernNOSPAMPLEASE.ch Link coverage report
SUSYTools atlas-phys-susy-backgroundforum@cernNOSPAMPLEASE.ch Link
CP Tools
General information about how to configure CP Tools can be found here: https://twiki.cern.ch/twiki/bin/view/AtlasProtected/PhysicsAnalysisWorkBookRel21CPRec

Object Definitions

Jets
Container name: AntiKt4EMTopo
Selection cuts: JetMC15
  • pT>25GeV
  • abs(η)<2.5
  • B-tagging: MV2c10 at 70% WP

Electrons
Container name: Electrons
Selection cuts: ElectronLikelihoodMC15

  • pT>25GeV
  • abs(η)<2.47
  • Veto 1.37<abs(η)<1.52
  • Loose (baseline): Identification: MediumLH, Isolation: no isolation
  • Tight (signal): Identification: TightLLH, Isolation: FixedCutTightTrackOnly

Muons
Container name: Muons
Selection cuts: MuonMC15

  • pT>25GeV
  • abs(η)<2.5
  • Loose (baseline): Identification: Medium, Isolation: no isolation
  • Tight (signal): Identification: Medium, Isolation: FixedCutTightTrackOnly

MET
Selection cuts:

  • Track Soft Term (TST), no fJVT
  • All Calibrated jets
  • electrons and muons passing loose selections
  • No photons and taus

Large-R Jets
Container name: AntiKt10LCTopoTrimmedPtFrac5SmallR20Jets
Selection cuts: JetMC15

  • pT>25GeV
  • abs(η)<2.0

Overlap Removals
Working Point: Standard

  1. remove electrons that shares track with muons
  2. remove single jet closest to an electron (within ∆R < 0.2)
  3. remove electrons/mouns within ∆R < 0.4 of any jet

Input Sample

TOPQ1 Powheg+Pythia8 ttbar MC sample
SCOPE:NAME GUID ADLER32 FILESIZE EVENTS
mc16_13TeV:DAOD_TOPQ1.12453997._001723.pool.root.1 F1CEF10F-5BB8-5045-B4E3-46503FED3B79 ad:8aafca9e 1.007 GB 6176

Results

Common Tools Configurations

Utilizing the asg::ToolStore::dumpToolConfig() function, CP Tool Configurations as dependent variables of Setups are studied.
Full List can be found here: https://docs.google.com/spreadsheets/d/14Lgpz19N_2ilQedNxoZ74kUp63hEJKLZLpqrEvab0-g/edit?usp=sharing
In Summary, most of the Tools Configurations can be easily configured and some hard-to-find discrepancies usually happens when CDI files of certain framework are not up-to-dated.

Cutflows & Yields

Cutflows

e+jets
Selection Nevents (AT) Nevents (ST) NevST-NevAT
INITIAL 6176 6176 0
Ne(pT > 25 GeV) ≥ 1 2144 2145 1
Ne(pT > 30 GeV) ≥ 1 1962 1963 1
Ne(pT > 30 GeV) = 1 1903 1904 1
Nμ(pT > 30 GeV) = 0 1776 1777 1
Pass jet cleaning 1776 1777 1
ETmiss > 20 GeV 1627 1627 0
ETmiss+mTW >60GeV 1532 1533 1
Njet ≥ 1 1528 1529 1
Njet ≥ 2 1448 1449 1
Njet ≥ 3 1193 1194 1
Njet ≥ 4 760 760 0
Nbjet ≥ 1 649 649 0
NlargeR-jet ≥ 1 157 157 0
Nbad-jet μ = 0 157 157 0
   
μ+jets
Selection Nevents (AT) Nevents (ST) NevST-NevAT
INITIAL 6176 6176 0
Nμ(pT > 25 GeV) ≥ 1 2478 2478 0
Nμ(pT > 30 GeV) ≥ 1 2227 2227 0
Nμ(pT > 30 GeV) = 1 2152 2152 0
Ne(pT > 30 GeV) = 0 2025 2025 0
Pass jet cleaning 2025 2025 0
ETmiss > 20 GeV 1842 1842 0
ETmiss+mTW >60GeV 1725 1726 1
Njet ≥ 1 1721 1722 1
Njet ≥ 2 1663 1664 1
Njet ≥ 3 1394 1394 0
Njet ≥ 4 922 922 0
Nbjet ≥ 1 794 794 0
NlargeR-jet ≥ 1 138 138 0
Nbad-jet μ = 0 138 138 0
weights Nevents (AT) Nevents (ST) NevST/NevAT
unweighted 157 157 100.0000%
weight mc 115589.492188 115589.492188 100.0000%
weight lep 153.322937 153.322937 100.0000%
weight btag 154.620438 154.620438 100.0000%
weight pu 156.81871 156.81897 100.0002%
weight jvt 156.280106 152.133163 97.3465%
All weights
   
weights Nevents (AT) Nevents (ST) NevST/NevAT
unweighted 138 138 100.0000%
weight mc 98668.734375 98668.734375 100.0000%
weight lep 136.900772 136.900772 100.0000%
weight btag 137.132629 137.132629 100.0000%
weight pu 123.823524 123.823769 100.0002%
weight jvt 136.602066 133.905701 98.0261%
All weights

Event-by-Event Comparison




Discussions

JVT (In)Efficiency Scale Factors TODO

Purple tag open issue Purple tag not confirmed
A significant discrepancy in the JVT scale factors obtained from the AnalysisTop and SUSYTools based frameworks is observed here. The JVT weight distributions of both is shown below:
JVTWeight.png
It is later understood that this "long tail" above 1.0 from AnalysisTop is featured by the JVT inefficiency SF. The current understanding is that because Efficiency scale factors are to be applied to hard scatter (HS) jets passing JVT, inefficiency scaling factors are to be applied to HS jets failing JVT. However, while OverlapRemoval itself is an input to the JVT inefficiency SF calculation, OverlapRemoval has to be done after the JVT cut. Before, this had caused an process issue in JVT SF calculation until the priority decorator feature was introduced. To conclude, today, instead of being thrown away directly, JVT-faling jets will be ignored by the OR Tool.

Given this historical factor, the survey now goes in the direction of checking if this is also implemented accordingly in SUSYTools. More details can be found in the discussion thread with AnalysisTop developers Link and their presentation of the statement of intent for that time Link.

-- YuHengChen - 2018-07-27

Edit | Attach | Watch | Print version | History: r24 < r23 < r22 < r21 < r20 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r24 - 2018-09-14 - HQTSoftwareTaskforceService
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Sandbox All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback