evidential_mse_loss¶
- probly.losses.torch.evidential_mse_loss(alphas: Tensor, targets: Tensor) Tensor[source]¶
Evidential mean squared error loss from [SKK18].
Implements the evidential MSE loss for classification uncertainty estimation, combining prediction error and predictive variance under a Dirichlet distribution.
- Parameters:
alphas – Dirichlet concentration parameters, shape (B, C).
targets – Ground-truth class labels, shape (B,).
- Returns:
Scalar evidential mean squared error loss averaged over the batch.
