evidential_kl_divergence¶
- probly.losses.torch.evidential_kl_divergence(alphas: Tensor, targets: Tensor) Tensor[source]¶
Evidential KL divergence regularizer from [SKK18].
Implements the KL divergence regularization term for classification uncertainty estimation in Evidential Deep Learning.
- Parameters:
alphas – Dirichlet concentration parameters, shape (B, C).
targets – Ground-truth class labels, shape (B,).
- Returns:
Scalar evidential KL divergence loss averaged over the batch.
