Integration of the ML based online Ecal DQM using autoencoder network

[abhih1]

PR description:

This PR introduces a machine learning based anomaly detection feature to the online ECAL DQM for the Ecal Barrel through a new DQM Client MLClient. An Autoencoder (AE) model with a ResNet architecture is trained on certified good data (EB digi occupancy) from 2018 runs.
On giving an input occupancy map to the AE, the encoder part of the AE encodes and learns the features and the decoder reconstructs the data from the encoded latent space to match the input as closely as possible. The reconstruction loss is then calculated, which is a mean squared error (MSE) between the input and output images at a tower level. Thus given an anomalous tower, the AE which has learnt the features of the good data will have a hard time reconstructing it and give a higher loss on the anomaly than on the good towers. A quality threshold is then applied on this loss map which marks it as Good or Bad, which is then stored as an ML quality summary plot.

We also do some pre processing and post processing steps to have a uniform, robust anomaly detection with fewer false positives over the changing detector conditions. The details of this are included in the comments throughout the code, and also can be found in this talk presented at a central DQM meeting [1]

Please note that the ML quality summary plot is a complimentary plot in addition to the existing DQM plots, and will be used for ECAL internal consumption only pending successful commissioning.
Along with the main ML quality summary plot, we also introduce a few monitoring plots to monitor the trends in Pileup (PU) per lumisection, no.of events processed in a lumisection and no.of events per input image to the ML model. This is in anticipation of the low luminosity expected during the start of Run 3. A few hyper parameters of our pre and post processing steps will also have to be tuned according to these trends, as currently they are derived from the conditions of Run 2.


Please also note that this version of the PR is only meant for code integrations tests during the November 2021 MWGR. The plots will only be meaningful when we have real collisions, and modifications are foreseen after gathering data from the beginning of Run 3.

[1] https://indico.cern.ch/event/1042141/contributions/4377869/attachments/2254007/3824216/ML4DQM_AbhiramiHarilal_May28.pdf

PR validation:

The code was validated by running the modified online Ecal DQM configuration with the new client enabled and the resultant plots were examined by uploading the output file to a DQM test gui.
The new plots are confirmed and look reasonable.

if this PR is a backport please specify the original PR and why you need to backport that PR:

N/A

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-35990/26420

• This PR adds an extra 3364KB to repository

[cmsbuild]

A new Pull Request was created by @abhih1 (Abhirami Harilal) for master.

It involves the following packages:

• DQM/EcalMonitorClient (dqm)
• DQM/EcalMonitorTasks (dqm)
• DQM/Integration (dqm)

@emanueleusai, @ahmad3213, @cmsbuild, @jfernan2, @pmandrik, @pbo0, @rvenditti can you please review it and eventually sign? Thanks.
@batinkov, @battibass, @rchatter, @argiro, @thomreis, @simonepigazzini, @threus, @francescobrivio 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

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-35990/26442

• This PR adds an extra 40KB to repository

[cmsbuild]

Pull request #35990 was updated. @emanueleusai, @ahmad3213, @cmsbuild, @jfernan2, @pmandrik, @pbo0, @rvenditti can you please check and sign again.

[cmsbuild]

-code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-35990/26443

• This PR adds an extra 40KB to repository

Code check has found code style and quality issues which could be resolved by applying following patch(s)

• code-format:
  https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-35990/26443/code-format.patch
  e.g. curl -k https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-35990/26443/code-format.patch | patch -p1
  You can also run scram build code-format to apply code format directly

[cmsbuild]

Pull request #35990 was updated. @emanueleusai, @ahmad3213, @cmsbuild, @jfernan2, @pmandrik, @pbo0, @rvenditti can you please check and sign again.

[jfernan2]

please test

[cmsbuild]

-1

Failed Tests: RelVals
Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-1461c9/20391/summary.html
COMMIT: 61c60ff
CMSSW: CMSSW_12_2_X_2021-11-08-2300/slc7_amd64_gcc900
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/35990/20391/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:

• @cms-sw Include tensorflow dependencies cmsdist#7429
• @mkirsano upgrade lhapdf to 6.4.0, sets to 6.4.0b cmsdist#7357

You can see more details here:
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-1461c9/20391/git-recent-commits.json
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-1461c9/20391/git-merge-result

RelVals

• 11634.911

----- Begin Fatal Exception 09-Nov-2021 18:14:40 CET-----------------------
An exception of category 'PluginNotFound' occurred while
   [0] Constructing the EventProcessor
Exception Message:
Unable to find plugin 'EcalBarrelGeometryEPdd4hep' in category 'CMS EDM Framework ParameterSet Description'. Please check spelling of name.
----- End Fatal Exception -------------------------------------------------

[jfernan2]

@cmsbuild please test with #36021

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-1461c9/20412/summary.html
COMMIT: 61c60ff
CMSSW: CMSSW_12_2_X_2021-11-09-1100/slc7_amd64_gcc900
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/35990/20412/install.sh to create a dev area with all the needed externals and cmssw changes.

Comparison Summary

Summary:

• No significant changes to the logs found
• Reco comparison results: 0 differences found in the comparisons
• DQMHistoTests: Total files compared: 42
• DQMHistoTests: Total histograms compared: 2901890
• DQMHistoTests: Total failures: 0
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 2901868
• DQMHistoTests: Total skipped: 22
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 41 files compared)
• Checked 177 log files, 37 edm output root files, 42 DQM output files
• TriggerResults: no differences found

[jfernan2]

+1
Backport 12_0_X Tested in Online DQM at P5

[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)

[qliphy]

+1