Privugger Measures

mutual_information

Mutual information estimators

privugger.measures.mutual_information.mi_sklearn(trace, var_names=[], disc_features=False, log2=True, n_neigh=20, input_inferencedata=True)[source]

Binding for estimating mutual information using the mutual_info_regression function from sklearn.

Concretely, this function computes mutual information as I(var_names[0]; var_names[1]).

Parameters

  • trace (az.InferenceData object) – InferenceData object containing the samples for variables of interest

  • var_names (String imes String) – String array with two elements indicating the variables used to compute mutual information. Note that the array must contain exactly two elements.

  • disc_features (bool) – Boolean indicating whether the domain of the secret is discerete. Default False.

  • log2 (bool) – Result in log2. Default True. If False, the result is in natural logarithm.

  • n_neigh (int) – Number of neighbours used by de estimator. Default 20.

Returns

result – Mutual information in log2 or ln.

Return type

float