incerto.data.LabelNoiseDataset

incerto.data.LabelNoiseDataset#

class incerto.data.LabelNoiseDataset(dataset, noise_rate=0.1, num_classes=None, seed=42)[source]#

Bases: Dataset

Add label noise to a dataset.

Useful for studying robustness to noisy labels.

Parameters:
__init__(dataset, noise_rate=0.1, num_classes=None, seed=42)[source]#

Initialize label noise dataset.

Parameters:

  • dataset (Dataset) – Base dataset

  • noise_rate (float) – Fraction of labels to corrupt (0-1)

  • num_classes (Optional[int]) – Number of classes (auto-detected if None)

  • seed (int) – Random seed

Methods

__init__(dataset[, noise_rate, num_classes, ...])

Initialize label noise dataset.
__init__(dataset, noise_rate=0.1, num_classes=None, seed=42)[source]#

Initialize label noise dataset.

Parameters:

  • dataset (Dataset) – Base dataset

  • noise_rate (float) – Fraction of labels to corrupt (0-1)

  • num_classes (Optional[int]) – Number of classes (auto-detected if None)

  • seed (int) – Random seed

Contents