bqlearn.corruptions.make_sampling_biais

bqlearn.corruptions.make_sampling_biais(X, *arrays, a=3, b=8, random_state=None)

Synthetic covariate shift by creating a sampling biais using the first axis of a PCA learned of the input features [1].

Parameters:

Xarray-like of shape (n_samples, n_features)

The samples.

*arrays: sequence of indexables with length / shape[0] equals to n_samples

Allowed inputs are lists, numpy arrays, scipy-sparse matrices or pandas dataframes.

afloat, default = 3.0

bfloat, default = 8.0

random_stateint or RandomState, default=None

Controls the randomness of the PCA.

Returns:

X_corruptedndarray of shape (n_samples, n_features)

The corrupted samples.

*arrays_imbalancedlist, length=len(arrays)

The corresponding imbalanced arrays.

References

[1]

Gretton, Arthur, et al. “Covariate shift by kernel mean matching.” Dataset shift in machine learning 3.4 (2009): 5.