Preparation of PO&DAS23

Thursday 29 Jun 2023, 10:00 → 12:00 Europe/Berlin

SR3a (Hamburg)

Zoom link: https://cern.zoom.us/j/61532761187?pwd=bnd0aFIrM3JHV05oYnpOZm11cFh6dz09
Meeting ID: 615 3276 1187
Passcode: 770451
 

Questions to all facilitators:

1. do have special computing requirements? (default: standard NAF account with NFS and HTCondor)
2. are you looking for more facilitators? (we want at least 2, ideally one of them being based in Hamburg)
3. are you planning any pre-exercises? (recommended to set up environment)
4. does your exercise depend on another exercise? (e.g. PAG inclusive jet depends on POG jets)
5. anything else?

Notes:

• We would like to finalise the list of exercises by mid of July.
• We are still looking for facilitators for the topical exercises.

1 Introduction

Speakers: Alexander Grohsjean (CMS - University of Hamburg), Andreas Hinzmann (CMS (CMS Fachgruppe Searches)), Matthias Komm (CMS (CMS Fachgruppe Searches)), Patrick Connor (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik) & CDCS)

plenary-talks.png

• Ankita: stat?
  • Patrick: not sure we have time... perhaps a bit in CAT for combine
• Matteo: "what to survive in an analysis" at CMS DAS @ CERN
  • Patrick: fits pretty well the last session

2 Topical exercise: Anomaly detection

Speakers: Louis Moureaux (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Patrick Connor (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik) & CDCS)

link

Abstract: This exercise is an introduction to novel techniques used to search for unusual occurrences in a dataset, for instance a few new physics events in an overwhelming standard model background. The two big classes of techniques will be presented.

Prerequisites: Basic knowledge of Python and machine learning will be assumed. The practical part of the exercise is based on PyTorch. It uses a Jupyter notebook that runs on Google Colab, locally, or at any supporting cluster (e.g. NAF or lxplus). Access to a CUDA-enabled GPU (as provided by Colab) helps but is not strictly necessary.

3 Topical exercise: Scikit Learn

Speakers: Laurids Jeppe (CMS (CMS Fachgruppe TOP)), Matthias Komm (CMS (CMS Fachgruppe Searches))

link

4 Topical exercise: RDataFrame

Speaker: Suvankar Chowdhury

5 Topical exercise: Git

Speakers: Lars Olaf Stietz (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Samuel Bein (None)

6 Topical exercise: MC techniques

Speakers: Mikel Mendizabal Morentin (DESY), Paris Gianneios, Polidamas Kosmoglou

7 Topical exercise: ROOT

8 Topical exercise: Coffea

9 Topical exercise: ML (Keras)

Speakers: Karim El-Morabit (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Lisa Benato (Hamburg University)

10 Topical exercise: Combine

Speaker: Aliya Nigamova (University of Hamburg)

The goal: learn how to setup different types of statistical models in the Combine Tool, the most common software for statistical inference in CMS, will also include a short session on debugging. 

Prerequisites: basic knowledge of python and the concept of fitting. 

11 Topical exercise: unfolding

Speakers: Matteo Defranchis (CERN), Patrick Connor (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik) & CDCS)

12 Topical exercise: ML (PyTorch)

13 Topical exercise: CMSSW

14 Topical exercise: ColumnFlow

15 DPG exercise: Trigger

Speakers: Artur Lobanov (Universität Hamburg), Juliette Alimena (CMS (CMS Fachgruppe Searches)), Sanu Varghese

abstractForTriggerExercise.rtf

TriggerExercise_PODAS2023_29June2023.pdf

• Sanu is STEAM convener, unclear if he will be in Hamburg for the school itself
• Which columnar tool in particular should be a good choice? will students already know about columnar analysis?
  •  it might be a requirement to take one of the columnar exercises before then
• Can we use it a local setup instead of SWAN?
  • to be followed up

16 DPG exercise: Tracker

Speakers: Dr Andrea Cardini (CMS (CMS Fachgruppe HIGGS)), Antonio Vagnerini (DESY), Henriette Petersen (CMS (CMS Fachgruppe TOP)), Sandra Consuegra Rodriguez (CMS (CMS Fachgruppe HIGGS)), Suvankar Chowdhury

• Pixel, strip, alignment, TkDQM
  • focus will be on TkDQM and alignment during the exercises (1h each)
  • focus on pixel and strip during the plenary
• alignment: give a feeling of how it works
• TkDQM: how to decide if a run is good or not
• will recycle material from the Tracker training days
  • includes simple hands-on exercise on alignment
  • ideas on TkDQM

17 DPG exercise: Calorimeters

Speaker: Fabrice Couderc

18 DPG exercise: Muon chambers

Speaker: Archie Sharma

19 POG exercise: PPD

Speakers: Louis Moureaux (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Patrick Connor (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik) & CDCS)

link

Please change the name to be more specific!

Abstract: In this exercise you will learn to find the samples you need for an analysis and information about them, for instance their cross section or how they were produced. You will also learn how the CMS simulation chain works and how to run it on your own.

Prerequisites: The exercise involves running CMSSW on the command line. It requires a working grid proxy.

20 POG exercise: Generators

Speakers: Dominic Stafford (CMS (CMS Fachgruppe TOP)), Luis Ignacio Estevez Banos (CMS (CMS Fachgruppe QCD)), Simone Amoroso (CMS (CMS Fachgruppe QCD))

POSDAS_generators_290623.pdf

• Patrick: Rivet is a must for this school, may be an extra exercise if you like, or part of the GEN exercise
  • Dominic: will think about it
  • Patrick & Andreas: we find it better standalone, but we let you facilitators decide
• Patrick: talk to Mikel, who will be setting up the exercise on MC techniques and application to QCD
• Matthias: Gridpack generation?
  • yes
• Matthias: MCM
  • covered by PPD
• Andreas: and there will be plenary talks

21 POG exercise: BTV

Speakers: Matteo Defranchis (CERN), Sebastian Wuchterl (DESY - CMS)

• Brand-new BTV exercise was used for CMSDAS @CERN (prepared by Annika Stein)
• Current version of the exercise can be complited in 2/3h, depending on the student
• Plan to add one/two more examples to make sure 4h are covered in all cases
• Trying to recruit person power for preparing the exercise and for facilitating. Will have a clearer idea in the next weeks

22 POG exercise: LUMI

Speaker: Joscha Knolle (Ghent University)

JK_230629_LumPogExercise.pdf

• Patrick: the exercise should be 4h long, I think you should cover BrilCalc
• Daniel: using hot data is really nice. But October is planned for HI, right?
• Josha: the exercise won't change much whether it's HI or pp. Otherwise I can identify a few interesting scans. On short notice, we can identify what we like to analyse.

23 POG exercise: CAT

Speakers: Aliya Nigamova (University of Hamburg), Daniel Savoiu (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Marcel Rieger (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Philip Daniel Keicher (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik))

• Daniel: planning to get together soon
• Laurids: topic?
  • job management & columnar analysis, in particular with "law"
  • HTCondor
• Patrick: requirements?
  • unclear at this stage
• PAtrick: we should set up a meeting with all computing exercises together 

24 POG exercise: Tracking

Speaker: Kirill Skovpen (Ghent University)

25 POG exercise: Muons

26 POG exercise: MET

27 POG exercise: Jet

Speakers: Andreas Hinzmann (CMS (CMS Fachgruppe Searches)), Anna Monika Albrecht (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Polidamas Kosmoglou

28 POG exercise: EGM

Speakers: Afiq Aizuddin Anuar (CMS (CMS Fachgruppe TOP)), Anshul Kapoor, Dr Ying An (CMS (CMS Fachgruppe TOP))

PODAS_EGamma_prepare.pdf

 

EGamma Exercise: (Jupyter notebook, mainly Python)

• Reconstruction/Calibration
• Energy correction
• Identification
• Tag & Probe method & MVA ID training

29 POG exercise: Proton

Speakers: Beatriz Ribeiro Lopes (CMS (CMS Fachgruppe TOP)), Michael Pitt

The forward proton detectors are new detectors introduced during LHC Run 2 by the CMS collaboration and are used to recontsruct scattered protons that emerge intact from the proton-proton interaction. In this exercise, we will learn about the process of proton reconstruction and its application in analysis, following the latest recommendations provided by the Proton Object Group.

Throughout this exercise, participants will use the nanoAOD data/MC samples with CMSSW, obtain analysis ntuple with proton objects, and make validation plots from the obtained ntuples using Jupiter notebook.

30 POG exercise: Taus

Speaker: Dr Andrea Cardini (CMS (CMS Fachgruppe HIGGS))

Previous plan

31 PAG exercise: Top mass measurement

Speakers: Henriette Petersen (CMS (CMS Fachgruppe TOP)), Matteo Defranchis (CERN)

The exercise is a bit old, but it was quite successful at the CMSDAS @CERN

• The minimal version of the exercise does not involve advanced analysis techniques
  • if the students are not very fast, they can anyway obtain a result
  • if the group is fast and efficient, there is a lot of margin for initiative with more modern techniques
    • for example, at CERN they implemented a fit in Combine from scratch, improving a lot on the baseline result
    • we also did the simultaneous unblinding of the two method, and used this as to cross-check the combine fit
       
• The exercise consists in producing control distributions, fitting the b-jet energy peak, calibrating the peak to the top mass, and estimating systematic uncertainties
• All the information is contained in simple flat ntuples, so the students can actually code the analysis without big technical obstacles
• The code is based on simple python scripts

 

UPDATE: technology for the exercise: simple python scripts, possibility to set up fit in Combine upon students' initiative. The fact that the code is very simple gives the student the possibility to do an analysis from beginning to end, without having to learn advanced software frameworks (which is very hard in just a few days)

32 PAG exercise: Long lived particle search

Speakers: Karim El-Morabit (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Lisa Benato (Hamburg University)

33 PAG exercise: BSM resonances in Run 3 (TBC)

Speaker: Daniel Savoiu (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik))

34 PAG exercise: B2G B*->tW search

Speaker: Matthias Komm (CMS (CMS Fachgruppe Searches))

35 PAG exercise: Higgs with columnar analysis

Speakers: Daniel Savoiu (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Philip Keicher (Universität Hamburg)

link

• Daniel: more concrete plans in the coming weeks

36 PAG exercise: TTbar cross section at 13.6 TeV

Speakers: Evan Ranken (CMS (CMS Fachgruppe TOP)), Laurids Jeppe (CMS (CMS Fachgruppe TOP))

PODAS meeting 290623.pdf

• Patrick: HTCondor will be available, but 1) prepare plan B, 2) make sure that students learn to test locally before
• Evan & Laurids: require exercise on Combine, probably one topical computing exercise
• Matteo: do you plan to use the same strategy as in the analysis, or central calib?
  • Laurids: probably the same strategy

37 PAG exercise: SMP Z->mumugamma

Speakers: Ankita Mehta (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Louis Moureaux (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik)), Paris Gianneios, Patrick Connor (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik) & CDCS), Dr Ying An (CMS (CMS Fachgruppe TOP))

link

Abstract: In this exercise you will measure a value missing from the PDG: the branching fraction of the Z boson to two muons and a photon. Starting from MiniAOD, you will perform the necessary corrections to the simulation and obtain a result free of detector effects.

Prerequisites: Ideally you should have followed the Muon and/or EGamma exercises. An understanding of generators and triggers will also be useful. The exercise will be done in C++ using the ROOT libraries.

38 PAG exercise: Inclusive jet

Speakers: Patrick Connor (UNI/EXP (Uni Hamburg, Institut fur Experimentalphysik) & CDCS), Polidamas Kosmoglou, Simone Amoroso (CMS (CMS Fachgruppe QCD)), Valentina Guglielmi (CMS (CMS Fachgruppe QCD))

link