Skip to content

Latest commit

 

History

History
60 lines (49 loc) · 2.65 KB

README.md

File metadata and controls

60 lines (49 loc) · 2.65 KB
Raw

Create input files for NN training

Before Usage

execute on NAF with CMSSW_9_4_9 or newer

uproot needs to be installed locally as it is not a part of the CMSSW package

pip install --user uproot

Adjust settings in preprocessing.py

  • base_selection to define a base event selection which is applied for all Samples

    (default base_selection = "(N_Jets >= 4 and N_BTagsM >= 3)")

  • change/add event categories (default event categories are ttH_categories and ttbar_categories)

	EVENTCATEGORYNAME=root2pandas.EventCategories()
  • change/add categories of event categories (SELECTION can be None)
   EVENTCATEGORYNAME.addCategory(CATEGORYNAME, selection = SELECTION)
  • ntuplesPath as absolute path to ntuples
  • memPath absolute path to MEMs, usage of MEMs is optional
  • change/add samples of the dataset used with
	dataset.addSample(SampleName  = SAMPLENAME,
    			ntuples     = PATHTONTUPLES,
    			categories  = EVENTCATEGORYNAME,
    			selections  = SELECTION,
    			MEMs        = PATHTOMEMS`
  • change additional_variables for variables needed to preprocess, that are not defined in the selection and not needed for training

Usage

To execute with default options use

python preprocessing.py

or use the following for options

  • -o DIR to change the name of the ouput directory, can be either a string or absolute path (default is InputFeatures)
  • -v FILE to change the variable Selection, if the file is in /variable_sets/ the name is sufficient, else the absolute path is needed (default is example_variables)
  • -e INT to change the maximal number of entries for each batch to restrict memory usage (default is 50000)
  • -n STR to change the naming of the output file
  • -m to activate using MEMs
python preprocessing.py -o DIR -v FILE -e INT -m -n STR

Concerning MEMs

In the current ntuple setup the MEM likelihood variables are not written directly to ntuples, but are saved separately.

To include MEM variables in the DNN training these need to be read and added separately. For the single-lepton analysis the MEMs are saved in .root files with trees called tree.

The function root2pandas.generateMEMdf reads the MEM values (more specifically mem_p) and event IDs (run, lumi, event) from these files and saves the information as separate pandas dataframes. At later steps in the preprocessing these MEM values are added to the output files. Only events where a MEM value was found (matching IDs) are added to the output files.

The MEM variable is saved as memDBp in the output files to match the naming scheme used in the single-lepton analysis.