xai4tsc.evaluation

XAI evaluation metrics wrapping the Quantus library.

Submodules:

  • evaluate — the evaluate() entry point and the METRICS registry.

  • base — the EvaluatorBase / QuantusEvaluator wrappers.

  • quantus_groups — what each Quantus metric consumes (QUANTUS_GROUPS) and the explanation domains that follow from it.

  • _ground_truth — the ground-truth region schema (GroundTruthRegion) and the masks the localization metrics score against. Datasets store plain mappings; this is where they are parsed.

  • frequency_evaluate, timefrequency_perturbation, timefrequency_auc — native frequency and time-frequency metrics.

Use evaluate() as the main entry point. The METRICS registry maps metric names to callables that produce an EvaluatorBase (native subclasses, or the single QuantusEvaluator bound to a name from QUANTUS_METRICS); use register_metric() to add custom ones.

Submodules