probly.train.evidential.torch.make_ood_target_alpha

probly.train.evidential.torch.make_ood_target_alpha(batch_size: int, num_classes: int = 10, alpha0: float = 10) → Tensor

Construct flat Dirichlet target distribution for out-of-distribution samples.

Used by Dirichlet Prior Networks, Posterior Networks, and PN-style paired losses to encourage high uncertainty on out-of-distribution inputs by assigning uniform Dirichlet concentration parameters.

Parameters:

• batch_size – Number of out-of-distribution samples in the batch.

• num_classes – Number of classes. Defaults to 10.

• alpha0 – Total Dirichlet concentration parameter (strength).

Returns:

Target Dirichlet concentration parameters, shape (B, C).