uncertainty/

Distribution-free prediction sets and intervals, probability calibration, and the metrics that verify both.

• Conformal prediction: SplitConformalClassifier,

SplitConformalRegressor, CVPlusRegressor, JackknifePlusRegressor, APSConformalClassifier, RAPSConformalClassifier, MondrianConformalClassifier, ConformalizedQuantileRegressor — prediction sets and intervals with a finite-sample coverage guarantee that needs no distributional assumption.

• Venn-Abers: VennAbersCalibrator — probability intervals that

report their own calibration uncertainty.

• Calibration: PlattCalibrator, IsotonicCalibrator,

TemperatureScaler, VectorScaler — post-processors that turn raw scores into probabilities you can act on.

• Metrics: coverage_score, average_set_size,

interval_width, brier_score, expected_calibration_error, maximum_calibration_error, reliability_curve.

These are post-processors around an already-fitted model rather than algorithms, so they are not registered in the TuiML algorithm hub.