API Reference¶

This section provides detailed documentation for all Calibre classes and functions.

Overview¶

Calibrating.

Calibration Methods — the calibration algorithms and base classes. Start with CenteredIsotonicCalibrator.
Cross-Validation — cross-validation shared by every calibrator, including cross_val_calibrate() for out-of-fold probabilities.

Measuring.

CORP Evaluation — CORP reliability diagrams and the MCB/DSC/UNC score decomposition. No bin count to choose, and none to tune in your favour.
Calibration Metrics — calibration error estimators, including the bias-aware ones and smECE, plus the granularity metrics.
Calibration Report — one call that gathers all of it, with optional bootstrap intervals.
Multiclass Calibration — class-wise evaluation and the diagnostic that tells you which multiclass method you need.
Plateau Diagnostics — where a fitted curve went flat, and how much data each plateau rests on.

Drawing.

Plotting — reliability diagrams, the score decomposition, and the resolution barcode. Needs pip install 'calibre[plots]'.

Supporting.