probly.train.evidential.torch.der_loss

probly.train.evidential.torch.der_loss(y: Tensor, mu: Tensor, kappa: Tensor, alpha: Tensor, beta: Tensor, lam: float = 0.01) → Tensor

Deep Evidential Regression loss for uncertainty-aware regression.

Combines a Student-t negative log-likelihood with an evidence regularization term as proposed by Amini et al. (2020).

Reference:

Amini et al., “Deep Evidential Regression”, NeurIPS 2020. https://arxiv.org/abs/1910.02600

Parameters:

• y – Ground-truth regression targets, shape (B,) or (B, 1).

• mu – Predicted mean of the Normal-Inverse-Gamma distribution, shape (B,).

• kappa – Predicted scaling parameter, shape (B,).

• alpha – Predicted shape parameter, shape (B,).

• beta – Predicted scale parameter, shape (B,).

• lam – Weight of the evidence regularization term.

Returns:

Scalar Deep Evidential Regression loss averaged over the batch.