MkFit TrackCandidate error rescale + relaxed selection for overlap hits

[mmasciov]

PR description:

This PR introduces two updates for mkFit tracking iterations:
(1) rescale covariance of TrackCandidate's before CMSSW final fit (where the rescaling option is configurable);
(2) relax selection of overlap hits.

It requires cms-data/RecoTracker-MkFit#9.

PR validation:

Rescaling the covariance of TrackCandidate's allows to better match CKF performance in terms of residuals, with a consequent decrease in duplicate tracks (from the duplicate merging/cleaning across mkFit and CKF iterations). E.g.:
http://uaf-10.t2.ucsd.edu/~mmasciov/MIC/rescaleError100/MTV_TTbarPU50_rescaleError100_defaultIters_mkFitOnly/plots_highPurity/effandfakePtEtaPhi.pdf

Similarly to CKF tracks, a few % bias in track pT is introduced, due to a known feature related the description of the detector material in endcaps. E.g.:
http://uaf-10.t2.ucsd.edu/~mmasciov/MIC/rescaleError100/MTV_TTbarPU50_rescaleError100_defaultIters_vsCKF/plots_highPurity/residualPt.pdf

Relaxing the selection of overlap hits allows to increase the track hit multiplicity, with a reduction of fake tracks:
http://uaf-10.t2.ucsd.edu/~mmasciov/MIC/rescaleError100/MTV_TTbarPU50_rescaleError100_extractedGeoms_overlapHitCuts_mkFitOnly/

For more details, please refer to presentations at TRK POG:
(1) https://indico.cern.ch/event/1133724/#4-mkfit-status-report
(2) https://indico.cern.ch/event/1142469/#5-mkfit-status-report

FYI @mmusich @vmariani

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-37580/29336

• This PR adds an extra 44KB to repository

[cmsbuild]

A new Pull Request was created by @mmasciov (Mario Masciovecchio) for master.

It involves the following packages:

• RecoTracker/MkFit (reconstruction)
• RecoTracker/MkFitCore (reconstruction)

@jpata, @cmsbuild, @clacaputo, @slava77 can you please review it and eventually sign? Thanks.
@makortel, @felicepantaleo, @GiacomoSguazzoni, @JanFSchulte, @rovere, @VinInn, @ebrondol, @gpetruc, @mmusich, @mtosi, @dgulhan this is something you requested to watch as well.
@perrotta, @dpiparo, @qliphy you are the release manager for this.

cms-bot commands are listed here

[slava77]

test parameters:

• pull_request = Update of mkFit JSON configurations following PR cms-sw/cmssw#37580 cms-data/RecoTracker-MkFit#9

[slava77]

@cmsbuild please test

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-e44227/23953/summary.html
COMMIT: 2158049
CMSSW: CMSSW_12_4_X_2022-04-15-1100/slc7_amd64_gcc10
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/37580/23953/install.sh to create a dev area with all the needed externals and cmssw changes.

The following merge commits were also included on top of IB + this PR after doing git cms-merge-topic:

• @mmusich Introduce PtMaxTrackCountFilter #37500
• @smuzaffar Fix CS8/EL8/CS9 openssl build error #37575

You can see more details here:
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-e44227/23953/git-recent-commits.json
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-e44227/23953/git-merge-result

Comparison Summary

Summary:

• No significant changes to the logs found
• Reco comparison results: 20071 differences found in the comparisons
• DQMHistoTests: Total files compared: 48
• DQMHistoTests: Total histograms compared: 3593159
• DQMHistoTests: Total failures: 37311
• DQMHistoTests: Total nulls: 1
• DQMHistoTests: Total successes: 3555825
• DQMHistoTests: Total skipped: 22
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.004 KiB( 47 files compared)
• DQMHistoSizes: changed ( 312.0 ): 0.004 KiB MessageLogger/Warnings
• Checked 200 log files, 45 edm output root files, 48 DQM output files
• TriggerResults: found differences in 5 / 47 workflows

[jpata]

What sort of changes do you expect downstream from tracking due to this change? As I understand, this is expected to reduce fake tracks, so would the following increase if PFCandidates be expected?

https://cmssdt.cern.ch/SDT/jenkins-artifacts/baseLineComparisons/CMSSW_12_4_X_2022-04-15-1100+e44227/49637/validateJR/all_OldVSNew_RunDoubleEG2017Cwf136p793/all_OldVSNew_RunDoubleEG2017Cwf136p793c_recoPFCandidates_particleFlow__reRECO_obj_eta.png

[mmasciov]

What sort of changes do you expect downstream from tracking due to this change? As I understand, this is expected to reduce fake tracks, so would the following increase if PFCandidates be expected?

https://cmssdt.cern.ch/SDT/jenkins-artifacts/baseLineComparisons/CMSSW_12_4_X_2022-04-15-1100+e44227/49637/validateJR/all_OldVSNew_RunDoubleEG2017Cwf136p793/all_OldVSNew_RunDoubleEG2017Cwf136p793c_recoPFCandidates_particleFlow__reRECO_obj_eta.png

@jpata, both the changes in this PR contribute to a slight increase of (true) tracks in the pseudorapidity region around 2.5 (see below). So, a similarly slight increase of PF candidates in that region should be expected. Being "good" tracks, then this should be considered a slight positive change.

In more detail:

• first, the newly introduced TrackCandidate error rescaling allows a slight increase of (true) tracks at |eta|~2.5:

• then, relaxing the selection of overlap hits on top of the error rescaling above allows for an additional slight increase in the same region (black vs. blue):

[mmusich]

type tracking

[jpata]

+reconstruction

• improved tracking performance around eta 2.5
• consequently, small reco changes expected in everything downstream from tracking
• requires Update of mkFit JSON configurations following PR cms-sw/cmssw#37580 cms-data/RecoTracker-MkFit#9

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @perrotta, @dpiparo, @qliphy (and backports should be raised in the release meeting by the corresponding L2)

[perrotta]

+1

• Changes observed in downstream quantities, e.g. MkFit TrackCandidate error rescale + relaxed selection for overlap hits #37580 (comment). are consistent with the explanation provided in MkFit TrackCandidate error rescale + relaxed selection for overlap hits #37580 (comment), and therefore expected (welcome, I would say...)