From the previous meeting:
"--> " indicate new things discussed during the meeting
1) DATA:
DONE- Filip shares features with Maxim.
DONE- Maxim email about he can download them
2) CODE, have it in a common repository:
DONE- Filip sent PR for merge
GitHub should have sent the notification request to the owners of the area.
--> (Virginia) will send a reminder, adding Fedor in cc
This will add Filip’s code in the https://github.com/yandexdataschool/cms-dqm/tree/master/notebooks area
Filip’s code means:
- feature plot overlapping code
- extractor root tree-> NumPyArray—> pickle
- Filip’s exercise on Auto-Encoder
3) FEATURES
DONE Filip added more variables/features to the analyzer (AOD -> root tree)
Added aod variables related to the main physics objects plus some variables, more machine related, that might benefit.
---> Confident that it's a pretty comprehensive panorama, we ask @all to have a look at them ans scream loud if somethinge evident is missed
Today: in his github area
- https://github.com/gnomeosaurus/Analyzers
- C++ code: https://github.com/gnomeosaurus/Analyzers/blob/AODanalyzer/AODAnalyzer/plugins/AODAnalyzer.cc
- config file for the c++ code: https://github.com/gnomeosaurus/Analyzers/blob/AODanalyzer/AODAnalyzer/test/AODAnalyzer_cfg.py
---> discussed further the opportunity to have it in a common repository as the rst of the code. Problem is that analyzer requires a CMSSW environment, so it might be difficult to just merge it with the notbook.
---> (Giovanni /Filip) will discuss technical details offline and will come back with ideas next time
4) twiki public with limitation to ml4dc egroup
DONE
https://twiki.cern.ch/twiki/bin/view/CMSML4DC
---> checked it is accessible and readable
---> (Virginia) will work on making it editable by the e-group member
5) data CERNBox of 5TB
TO be DONE Giovanni, Gianluca as CMS L1
please let Virginia know if help needed
---> Giovanni report that he does not believe it's something that requires PPD group superpowers.
---> (Giovanni) will either investigate if it is a "eos admin" matter or, in case it does not require any special superpower, he will send instruction to Virginia.
---> probably the access to this area will be granted to an e-group. Question: core team only or everybody?
6)
offline Run Registry (database and its web interface where CMS stores DC informations)
DONE : Virginia asked Valdas how to include the info
- in RR V3(today): possibility to add a column for the info. He wouldn’t say how to “write in it”
---> (Virginia) will re-iterate with Valdas about how technically we could write/test in 2017-2018 data taking
- in RR v4(2020): ML projects are already part of the RR upgrade project. an upgrade RR workshop will be in the fall, ML4DC group could (via Virginia) suggest preferred option to Valdas, in a way he will take them in consideration during the design.
7) correlation grid: Maxim output 31
??? understood
---> (Fedor) said he is looking into it and he will investigate some discrepancies of algo behaviour in relation to CSC detector.
he will follow up with some people at FNAL.
---> (Virginia) will look for DC documentation about this detector and send around in PDF format (because internal CMS twiki pages) if she finds something
——————
NEW items:
8)
new member of ml4DC core team: Lukas Danev, OpenLab summer student working with Filip on Reduction of variable and Giovanni as supervisor, At CERN until end of August.
9) talk of ml4dc status at next CMS Machine Learning Workshop https://indico.cern.ch/event/646801/overview
slides are attached
A.O.B