jax

JAX implementation of Metrics.

Note

There is no JaxSingletonCredalSet or JaxDiscreteCredalSet in probly.representation.credal_set.jax; for those semantics, use the numpy-side NumpySingletonCredalSet / NumpyDiscreteCredalSet types. The remaining jax credal sets (Convex, DistanceBased, ProbabilityIntervals, DirichletLevelSet) all use the interval-dominance rule via their lower() / upper() envelopes.

Functions

jax_categorical_log_loss_per_instance

Score categorical log probabilities directly, preserving extreme logits.

jax_classwise_ece

Compute the classwise expected calibration error for JAX arrays.

jax_expected_calibration_error

Compute the confidence expected calibration error for JAX arrays.

jax_log_loss_per_instance

Compute log loss without logging zero probabilities of classes with no target mass.

jax_log_probabilities

Compute the log probabilities of a JAX array.

jax_one_hot_argmax

One-hot encode the first maximum of each probability vector of a JAX array.

jax_precision_recall_curve

Compute precision-recall curve along the last axis.

jax_probabilities

Return a JAX array of probabilities unchanged.

jax_probability_categorical_log_loss_per_instance

Score normalized categorical probabilities with target-aware logarithms.

jax_roc_curve

Compute ROC curve along the last axis.