multiaug.augmenters.tabular package¶
Submodules¶
multiaug.augmenters.tabular.gaussian_perturbation module¶
-
class
multiaug.augmenters.tabular.gaussian_perturbation.
GaussianPerturbation
(method: str = 'variance', fraction: float = 0.1)¶ Bases:
multiaug.augmenters.meta.Augmenter
Apply featurewise Gaussian noise to each feature in a sample.
Parameters: - method (str) –
Method to use to determine the noise (currently variance is the only supported method).
- If ‘variance’ then the noise is determined by taking a fraction of the variance (across the individual features) for each feature and adding it the the original feature.
- fraction (float) – Fraction of noise to add to sample.
-
apply
(data: numpy.ndarray, row_ids: numpy.ndarray)¶ Apply the augmentation to the samples.
Parameters: - data (np.ndarray) – Entire dataset such that the noise can be determined relative to the entire dataset.
- row_ids (np.ndarray) – The indices of the samples to which to apply noise.
Returns: Augmented data
Return type: np.ndarray
- method (str) –