compute_metrics#

scio.eval.compute_metrics(confs_ind, confs_oods, metrics)[source]#

From precomputed confidence scores, evaluate scores.

Parameters:

confs_ind (NDArray) – First output of compute_confidence().
confs_oods (tuple[NDArray, ...]) – Second output of compute_confidence().
metrics (tuple[BaseDiscriminativePower, ...]) – The different types of metrics to compute for every (score, ood set) combination.

Returns:

evals (NDArray[np.floating]) – Discriminative Power for every possible combination of score, OoD set and metric. Shape corresponding to (n_scores, n_ood_sets, n_metrics).