research.active_learning.entropy

research.active_learning.entropy(unlabeled_ids, increment, probabilities, **kwargs)

Sample selection based on Entropy selection criterion.

Parameters

unlabeled_idsarray-like of shape (n_samples,)

Indices of the unlabeled samples in the original (unlabeled training) dataset.

incrementint

Number of observations to select.

probabilitiesarray-like of shape (n_samples, n_classes)

Class probabilities of the input samples belonging to the unlabeled dataset.

Returns

new_idsarray of shape (increment,)

Indices of unlabeles samples to be added to the labeled training dataset.