config_separator.yaml

#Configuration file for model training and application.

# seed for the random number generators, to make things reproducible
seed: 0
true_energy_column: corsika_event_header_total_energy

separator:
  #the classifier to use
  classifier : |
    ensemble.RandomForestClassifier(
        n_estimators=30,
        max_features='sqrt',
        n_jobs=-1,
        max_depth=15,
        criterion='entropy',
    )

  # randomly sample the data if you dont want to use the whole set
  n_background: 500
  n_signal: 500

  # define the number of cross validations to perform
  n_cross_validations : 5

  # Define the name of the output column for the positive class.
  # default is gamma_prediction
  output_name: gammaness

  # Define the name of the column that contains the name of the telescope in case you're working
  # with mulitple telescopes and telescope types
  # telescope_type_key: telescope:name

  features:
    - size
    - length
    - width
    - num_islands
    - leakage1
    - leakage2
    - skewness_long
    - skewness_trans
    - concentration_cog
    - concentration_core

  # Generate some features using pd.DataFrame.eval
  # List all columns that have to be read from the input files
  # in `needed_columns`.
  # features must be a mapping of feature name -> expression
  feature_generation:
    needed_columns:
      - width
      - length
    features:
      area: width * length * @pi