Bug fix for GED electron MVA preselection calculator

[lgray]

Properly initialize and use some variables that were missed in the calculation of the electron preselection-mva.

These variables were present in the training, but not in the evaluation, so they carry some rejection power.

Small expected changes for tracker-driven only electrons, small improvements in efficiency/fake-rate.

Now includes #2717 and #2719 .

We should backport this to 700.

[cmsbuild]

A new Pull Request was created by @lgray (Lindsey Gray) for CMSSW_7_1_X.

Bug fix for GED electron MVA preselection calculator

It involves the following packages:

RecoEgamma/ElectronIdentification

@nclopezo, @cmsbuild, @anton-a, @thspeer, @slava77, @Degano can you please review it and eventually sign? Thanks.
You can sign-off by replying to this message having '+1' in the first line of your reply.
You can reject by replying to this message having '-1' in the first line of your reply.
@nclopezo, @ktf you are the release manager for this.
You can merge this pull request by typing 'merge' in the first line of your comment.

[slava77]

Hi Lindsey, could you please post some comparison plots describing the changes here so that we roughly know what to expect in testing and, maybe more importantly, to have a reference for the release history

Thanks.

[lgray]

Hi Slava,
I'll post some of the plots from Simon here in a bit.
There's also going to be a separate PR for an(other) update to the preselection MVA.
Simon was able to improve the size and composition of the training sample.

[cmsbuild]

+1
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-2558/250/summary.html

[dbenedet]

Hi Slava, here the plots from Simon. The fake rate increase in the endcap but it is still below 1% that is our goal. At the same time we have un increase of efficiency.

[dbenedet]

@lgray we would need as usual to update the weight file in a second moment. Thanks.

[slava77]

Hi Daniele and Lindsey, please clarify on your comment about the weight file.
Were your plots made with the weight file updated or not?

The file is in the external: was the request for an update submitted?
Once it's updated, this PR would need to have a new download.url, I guess, unless you stay with the old name.

Thanks

[dbenedet]

Hi Slava,

yes the plots are done with the new weight file. However it change very
little, with respect to the old trained file.
The external file is available in the usual /afs directory but I think
that still it needs to be requested.
When it will be available we will update the download.url and the
corresponding python files.

cheers,
Daniele

On 27/02/14 01:07, slava77 wrote:

Hi Daniele and Lindsey, please clarify on your comment about the
weight file.
Were your plots made with the weight file updated or not?

The file is in the external: was the request for an update submitted?
Once it's updated, this PR would need to have a new download.url, I
guess, unless you stay with the old name.

Thanks

—
Reply to this email directly or view it on GitHub
#2558 (comment).

[slava77]

Hi Daniele,
Since the external file and download.url update will need another check, I just wanted to reduce the number of incremental epsilon tests and do it in one go.
What's the request number in cmsdist for this file update?
Please let me know

[slava77]

@lgray @dbenedet
Daniele and Lindsey, what's the status with the external update with the new weight file?

[slava77]

Hi Lindsey

Once cms-sw/cmsdist#465 is resolved and is in the IB,
I'd rather see all changes from #2717 and #2719 included here and the other two PRs closed.

Giulio, is it too late for 05-0200 IB?

[lgray]

Hi Slava,
No problem, once it's included I'll toss the other two commits in here.

[ktf]

It was too late for me ;-) @Degano can you update the files? Thanks.

[lgray]

@ktf @Degano Please let me know which IB the updated payload will go in, thanks!

[ghost]

Hi,
I've updated the files in cms-sw/cmsdist@7f4cfab
And merged the PR already, so it should go in this afternoon IB (1400).

[slava77]

Great. Thanks, Alessandro.
Lindsey, you can update this already (no need to wait until ~6-8PM for this to actually show up in scram)

[cmsbuild]

Pull request #2558 was updated. @nclopezo, @cmsbuild, @anton-a, @thspeer, @slava77, @Degano can you please check and sign again.

[nclopezo]

@lgray
Hi Lindsey, this pull request has become unmergeable, could you please rebase it?

[lgray]

@nclopezo now based on CMSSW_7_1_X_2014-03-05-0200

[cmsbuild]

Pull request #2558 was updated. @nclopezo, @cmsbuild, @anton-a, @thspeer, @slava77, @Degano can you please check and sign again.

[ktf]

@Degano the changes you made broke the IB. Please add back the files you removed (creating a new cms-data). We should remove files only when they are not used anymore.

[nclopezo]

+1
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-2558/388/summary.html

[slava77]

checking it now

[slava77]

+1

for #2558 2192eff
Changes as expected: all diffs start at the MVA (it peaks better at 1 for electron guns).

• Some increase in ged electrons in the endcaps in ele gun sample
• Around 10% increase in ged electron counts in the jetty/background-like samples.

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next CMSSW_7_1_X IBs unless changes (tests are also fine). @nclopezo, @ktf can you please take care of it?