User Tools

Site Tools


Sidebar

2015:groups:tools:lhaad:proposal

This is an old revision of the document!


Here we will work on a combined proposal using the other two by combining their strengths and eliminating their weaknesses. We first define the rules for each type of block and the structure of the file in total. We then follow with an explicit proposal that should correspond to some real-life analyses from ATLAS and CMS.

Open Questions/Comments

  1. D: Maybe the overall description comment should start with '###' as in Python to distinguish from normal '#' comments which the parser should just ignore
    • S: The proposal concerning '#' is a really tricky thing. How about including three single quotes like in python docstrings? or /* description*/ like in c++
  2. D: See 'detectorobj' description: A 'take external' selection could clearly point out those objects which have to be defined on a fundamental level in whatever program one uses
  3. D: What is the difference between 'doc' and 'code'? Can't the 'doc' contain everything, including the code?
    • N: 'doc' is a link to a human-readable document. 'code' is the name of the function/link to the function (for when they make a library of functions).
  4. D: What is the purpose of 'eff' in the cuts and what distinguishes a 'trigger' selection from a normal 'select'?
    • N: There is no difference between 'trigger' and 'cut'. In the original version, one could have aribitrary xml tags, so I thought it'd be nice to have explicit names. Now, everything should be a 'cut'.
    • N: The eff is an estimated efficiency for things like the trigger/b-tagging etc. provided by the experiments. We should in principle also allow inclusion of efficiency tables instead of a single number.
  5. D: In cuts, I again would rather use 'take' instead of 'contains': if 'B' uses the events from A, then 'A takes the events from B' (but 'B contains the events from A, as A applies more cuts')
    • N: Maybe 'include' instead of 'take'? The idea behind 'contains' was a generic inheritance (not just for cuts) where any block can inherit from any other.
  6. D: Is it possible to use '|' instead of .OR.? the dot to me always refers to a member of a class.
    • N: I don't see why not (or even a c-like || which is easier for a human to read than a single |).

General Structure

The idea is to write a plain text file that can be read easily into xml if needed (and hence universally readable). The name of the block becomes the xml tag. Anything that follows is turned into attributes. e.g. “name ATLAS” inside the block becomes <experiment name=“ATLAS”>. Anything inside the block starting with “#” is added as a comment, i.e. the block

info analysis
  # Details about experiment 
  id ATLAS-2014-XXX
  publication JHEP11(2014)118
  sqrtS 8.0
  lumi 20.0

converts to

<info name="analysis" id="ATLAS-2014-XXX" publication="JHEP11(2014)118" sqrtS="8.0" lumi="20.0" description="Details about analysis">

There are five types of pre-defined blocks:

  1. detectorobj (defines reconstructed objects)
  2. function (defines functions that act on these)
  3. cut (blocks of cuts)
  4. binning (following Daniel's idea, for complex descriptions of orthogonal signal regions)
  5. info (for meta information chosen by the implementer, e.g. information about publication, conf-notes, theory papers used for interpretation, benchmarks, etc.)

For presenting results, we need two more blocks, e.g.

  1. results (for expected, observed, S95, etc.)
  2. cutflow (for cutflows)
  3. table (for providing digitised histograms etc.)

For any block, a link to code or external documentation can be provided. We can also include a MCsamples block to document the programs used to generate signal and background samples.

info

info analysis
  # Details about experiment 
  id SUSY-2013-15
  publication JHEP11(2014)118
  sqrtS 8.0
  lumi 20.0
  arXiv 1407.0583
  hepdata https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PAPERS/SUSY-2013-15/
info units
  ### Details about units for dimensionful quantities
  energy GeV
  length mm
  xsec pb
  # If using function repository, need to clarify if phi is between 0-2pi or -pi to pi.

function

function function_name
  arg1 type-of-arg1
  arg2 type-of-arg2
  return type-of-returnvalue
  code link-to-actual-code
  doc link-to-documentation

Rules:

  • allowed keywords: argX, return, code, doc
  • argX must count from 1 upwards
  • it is recommended to follow with a comment about its meaning after each argument
  • allowed types: int, float, string, bool, detectorobj, tlorentzvector, X-list (X is any of the allowed types)
  • for type 'detectorobj' it should be clear from the documentation which information the detectorobj needs to have

Example:

function isolation
  ### Sums up activity in the vicinity of a given candidate
  arg1 detectorobj # a single electron, muon or photon
  arg2 string      # "calo", "tracks", "eflow"
  arg3 float       # dR cone to be probed
  arg4 ptmin       # minpt of objects to be counted
  arg5 bool        # divide by candidate's pt?
  return float # sum of pT of the respective surrounding objects
  code http://google.de
  doc http://google.de
 
function overlap
  ### Checks if the candidate overlaps with any object of the list provided
  arg1 detectorobj      # to be tested candidate
  arg2 detectorobj-list # to be tested potential neighbours
  arg3 float            # dR cone to be probed
  return bool           # true if any of the neighbours is closer than dR to the candidate
  code http://overlap.de
  doc http://overlap.de
 
function hasMediumTag
  ### Checks if the candidate overlaps with any object of the list provided
  arg1 detectorobj      # electron
  return bool           # true if it is tagged as medium
  code http://electrons.de/medium
  code http://electrons.de/medium

detectorobj

Rules:

  • Each detectorobj must start with the keyword “take X”. X can either be a detectorobj-list X defined before or 'external'. In that case, the list is a fundamental object to be defined globally.
  • select or reject keywords can filter out specific candidates out of the input list: They have to be followed by a Boolean expression, in which a single candidate of the given detectorobj list can be tested via name.? (see examples)
  • functions can access previously defined detectorobj-lists by their name, but not the current detectorobj-list itself
  • additional optional keywords can give more insightful information about commonly used parameters.

Example:

detectorobj muon_l
  ### Loose muons
  take external
  select isolation(muon_l.?, 'tracks', 0.4, 0.5)<0.1
  code http://muons.mu
  doc http://checkmate.muons.mu
 
detectorobj elec_l
  ### loose electrons
  take external
  select isolation(elec_l.?, 'tracks', 0.4, 0.5)<0.1
  reject overlap(elec_l.?, muon_l, 0.4)
  code http://elecs.el
  doc http://checkmate.elecs.el
 
detectorobj elec_m 
  ### medium electrons
  take elec_l
  select hasMediumTag(elec_l.?)
  code http://elecs.el/medium
  doc http://checkmate.elecs.el/medium
 
detectorobj lep 
  ### signal leptons
  # lep contains medium electrons and loose muons
  take elec_m 
  take muon_l
 
detectorobj jet
  ### clustered jets from the calorimeter cells
  take external
  algorithm anti-kt
  R 0.4
  ptmin 20
  etamax 2.5
  code http://jets.jet
  doc http://checkmate.jets.de

cut

Rules:

  • a cut block processes a list of conditions from top to bottom:
    • 'select <bool>' only lets events pass for which the boolean expression holds true
    • 'reject <bool>' only lets events pass for which the boolean expression holds false
  • eff?
  • trigger?
  • .OR.?
  • allowed brackets for functions?

Example:

cut preselect
  ### Pre-selection cuts
  eff 0.95
  trigger e_trigger.OR.mu_trigger.OR.met_trigger
  select lep.1.pt>25
  reject lep.size>1
  select jet.size>2
  select met>100

binning

results

cutflow

info MCsamples
  ### All information concerning signal and background generation goes here
  signal Sherpa1.4.1 stop-pair
  background POWHEG ttbar mtop = 172.5 NLO
  table maasinfo
  doc http://singalgenerator.generate
 

table

Examples

2015/groups/tools/lhaad/proposal.1435517569.txt.gz · Last modified: 2015/06/28 20:52 by suchita.kulkarni