User Tools

Site Tools



MC, Tools and ML

Systematic uncertainties [Thurs 14h00-16h00 Auditorium]

Slack channel: #np-ps-syst-uncs

Possible projects:

  • Shower accuracy and benchmarks
  • Amplitude evolution, if people are interested
  • Role and impact of the infrared cutoff, tuning, and uncertainties
  • A practical approach for experiments
    • What comparisons make sense - what is available?
    • Generators wishlist?
  • Including systematic uncertainties in ML based analysis

Interested people: Josh, Simon, Stefan, Vinny, Axel, Aishik, Saptaparna, Andrzej, Ana, Maria V., Gianna, Chris G, Christian P, Max, Peter M

Agreed projects:

  1. Perturbative benchmarks of showers [Simon, James, Daniel]
    1. NLL, Numerical impacts
  2. Interplay between shower and hadronisation [Simon]
    1. Extending study from LH17 arXiv:1803.07977
    2. Effective/comprehensive uncertainty band
  3. Uncertainty recommendation using available tools [Josh, Gianna]
    1. Setup 4-point variations of all combinations of shower and hadronisation model
    2. Tune these in a consistent way to get an uncertainty envelope
  4. Tuning strategies [Stefan K]
    1. What is currently done?
    2. Common framework?
    3. Better documentation of existing framework
    4. Benchmark sets of data

Generator software and computing considerations [Friday 11h00-12h30 Auditorium]

Slack channel: #gen-soft-comp

Possible projects:

  • Efficiencies
  • MC reweighting/unweighting
  • GPU/Vectorisation
  • Event format/sharing
  • Containerisation/preservation/reproducibility
  • Use of ML
  • Common frameworks

Interested people: Josh, Vinny, Andrea, Stefan, Saptaparna, Maria V., Sergio, Chris G, Christian P, Ana, Max, Peter M, Simon, Sudeepan

Agreed projects:

  1. GPU/Vectorisation/HPC [Josh]
    1. Experience porting codes
    2. Workflows in experiments
      1. How would code be run in practise
      2. How to get/confirm allocations
      3. Sharing of resources between EXP and TH
  2. Computing performance [Josh]
    1. Benchmarking current code
      1. Also with examples from EXP
    2. Accounting in experiments
    3. Projections for future N(2,3)LO calculations
  3. Updates to interfaces [Josh]
    1. Interoperability of models
    2. Modular framework
    3. Multi-event API in MC generators
      1. Resampling
  4. Accessibility of tools [Josh]
    1. Containerisation and reproducibility
      1. Docker images, run cards, Rivet routines, Yoda files → HepData
      2. Use these tools to make LH studies reproducible
      3. Can at Docker-based workflow on e.g. lxplus be documented to lower barrier to entry for MC studies?
      4. Make sure cards actually used by EXP are uploaded with TH prediction Yoda files.
      5. Make available for benchmarking (c.f. point 2.), including full chain in EXP
    2. MC Generation generic tool/interface
      1. Like MCPlots
  5. Event formats [Josh]
    1. New HDF5 standard ← LHEF/HepMC
      1. Easier event sharing, smaller disk footprint
    2. Need for more information in intermediate stages of generation?
    3. Sample sharing between experiments
      1. Joint/cross validation
  6. ML for MC []
    1. Interoperability link to point 3. (swapping out parts of generation in modular framework)
    2. Discussion of exact vs approximate use of ML
  7. Resampling/reweighting [Josh]
    1. Reweighting and derivative-based optimisation
    2. Resampling to eliminate negative weights
      1. NN and cell-based tools
      2. Stress tests of these tools?
        1. Problem areas from TH PoV
        2. Validation at scale in EXP

ttW modelling studies [TBC]

Slack channel: #ttw

Possible projects:

  • Understanding differences between Sherpa MEPS@NLO and aMC+Py8 FxFx?
  • Modelling uncertainties?

Interested people: Josh, Sergio, Stefan, Gianna

Issues with EW evolution, identification of EW final states [Thurs 16h00-17h30 QCD room]

Slack channel: #ew-evo-id

  • Possibly connecting to ttW?

Interested people: Simon, Axel, Andrzej

Interplay between showers and hadronization

  • Hadronization model should respond to shower IR cut variations
  • Study IR ctuoff dependence, map out how purely perturbative variations compare
  • goal is to update the LH 17 study on retuning with a more solid theory framework

Interested: Simon, Andrzej, Stefan K.

Tuning Accord and Strategies

  • What is currently done:
    1. Common framework for tuning exist, but no single script to run
      1. Nice to have a script, which automatically prints exactly what is done and with which flags
    2. Every generator team does tuning on their own
    3. Almost no documentation on the tunes provided
      1. Rarely publicly available weights
      2. Rarely publicly available tuning options (samples, interpolation order, flags used, …)
      3. Often post-processing tune by eye
  • Goals of the project:
    1. Coordination among generator authors of Herwig, Pythia and Sherpa (mattermost channel?)
    2. Common framework for tuning exist, but no single reproducible script:
      1. Stefan K. trying to provide this in Professor, which automatically documents what version, options, flags, etc. are being run
      2. Someone could do something similar in Apprentice
    3. Les Houches Accord on some Benchmark observable+weight sets:
      1. Observable quality & quantity requirements:
        1. No ratio Rivets if full information is available?
        2. Reduce double counting of observables:
          1. No gaps in histogram (see e.g. some ALEPH identified particle momentum distribution)
          2. Smaller bin-width is preferred
          3. Enough statistics:
            1. NOTE: If the errors of the data are to small a theory uncertainty should be provided to the tuning framework otherwise the minimization will likely over-fit this data
      2. Benchmark sets:
        1. Conservative set of selected (IRC safe?) Event Shapes:
          1. Input from perturbative experts needed
          2. Only close to the peak, where hadronization is not dominant
        2. Provide global sets from each generator to gather information
        3. Ultimate goals:
          1. Minimal set of accorded observables+weights among all generators for comparability
            1. Tune using $\chi^2$ (more strict) and $\chi^2/(1+\chi^2)$ (less strict)
          2. Maximal set of accorded observables+weights among all generators for comparability
            1. Tune using $\chi^2$ (more strict) and $\chi^2/(1+\chi^2)$ (less strict)
    4. Provide recommendations to experiments on tuning?

Interested: Simon, Andrzej, Stefan K., ….

Status of q/g discrimination [Weds 14h00-15h30 Library]

  • Linked to Jets and Higgs

Interested people: Simon, Andrzej

Fiducial multiboson cross sections (VBF/VBS) [Thurs 11h00-12h30 QCD room]

  • Linked to Higgs

Interested people: Simon, Josh, Stefan, Axel, Saptaparna

Possible HF review [TBC]

  • Linked with Higgs and Jets

Interested people: Josh, Davide?, Stefan

ML-focused [TBC]

Possible projects:

  • New unfolding observables
  • Interoperability/practical considerations

Interested people: Vinny,Josh, Aishik, Andrzej, Maria V., Peter M

Overlaps between FPF and LHC [TBC]

Slack channel: #lhc-fpf

(Possibly the same for DUNE and EIC if enough people are here?)

Possible projects:

  • Impact of NNPDF + FPF data on LHC precision, e.g. W-mass

Interested people: Josh, Maria V., Stefan, Peter M (EIC)

2023/groups/smtools/start.txt · Last modified: 2023/06/19 11:11 by stefan.kiebacher