ml

This module contains the functions needed to train a Machine Learning-based Montecarlo scan, and to assess its performance.

SMEFT19.ml.SHAP_bf(fmodel, bf)[source]

Computes the SHAP values of the best fit point

Arguments

  • fmodel: Path to the file where the model was saved.

  • bf: Best fit point.

SMEFT19.ml.SHAP_param(fmodel, points, param, header=None)[source]

Creates an scatter plot displaying how the SHAP values change as functions of each parameter of the fit.

Arguments

  • fmodel: Path to the file where the model was saved.

  • points: Pandas Dataframe containing the dataset.

  • param: Fit parameter. 0 = C, 1 = al, 2 = bl, 3 = aq, 4 = bq.

  • header: If the data file contains headers in the first row, 0.

SMEFT19.ml.SHAP_summary(fmodel, points, fout, header=None)[source]

Creates a summary plot of the average SHAP values on a dataset.

Arguments

  • fmodel: Path to the file where the model was saved.

  • points: Pandas Dataframe containing the dataset.

  • fout: Path to save the plot (pdf only).

  • header: If the data file contains headers in the first row, 0.

SMEFT19.ml.hist(ML, vpoints, fout)[source]

Plots an histogram for the predicted and actual likelihoods, and compares them to the chi-square distribution

Arguments

  • ML:The Machine Learning scan module.

  • vpoints: Path to the file containing the points in the validation dataset.

  • fout: Path to save the histogram.

SMEFT19.ml.lh(x)[source]

Pickle-able function for the likelihood in scenario BII.

SMEFT19.ml.load_model(fmodel, vpoints, bf)[source]

Loads a XGBoost model previously saved

Arguments

  • fmodel: Path to the file where the model was saved.

  • bf: Best fit point.

Returns

  • Machine Learning scan.

SMEFT19.ml.regr(ML, vpoints, fout)[source]

Plots the predicted likelihod vs the actual likelihood and computes their regression coefficient

Arguments

  • ML:The Machine Learning scan module.

  • vpoints: Path to the file containing the points in the validation dataset.

  • fout: Path to the output regression plot (pdf only).

Returns

  • A tuple containing the Perason r coefficient and the p-value of the regression

SMEFT19.ml.train(dataset, fval, fmodel, bf, headers=None)[source]

Trains the Machine Learning algorithm with the previously computed Metropolis points

Arguments

  • dataset: Path to the file or list of files containing the Montecarlo pre-computed points.

  • fval: Path to the file where the validation points will be saved.

  • fmodel: Path to the file where the XGBoost model will be saved.

  • bf: Best fit point.

  • headers: Header lines in the dataset files. None if there is no header, 0 if the first line contains the header. Admits list if using several dataset files.

Returns

  • The Machine Learning scan module, already trained and ready to be used

previous

plots

next

utils