Pseudo Labeling

mlcpl.pseudo_labeling.CurriculumLabeling(*args, **kwds)

An implementation of curriculum labeling.

mlcpl.pseudo_labeling.CurriculumLabeling.__init__(self, dataset)

Args:

dataset: The 'MLCPLDataset that is applied to.

mlcpl.pseudo_labeling.CurriculumLabeling.update(self, model, batch_size=32, num_workers=20, selection_strategy='score', selection_threshold=0.5, verbose=False)

Update pseudo-labels.

Args:

model:

the model that produces predictions.

batch_size: size of mini-batches.

num_workers: the number of data loader workers.

selection_strategy: The method to select predictions to generate pseudo-labels.

selection_threshold: The thresholds for generating pseudo-labels.

verbose: Whether showing the process of generation.