🎯 Isotonic Regression

Isotonic Regression#

A nonparametric, monotonic fit used to calibrate predicted probabilities.

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

What it is#

Isotonic regression is a non-parametric calibration method that fits a monotonic (non-decreasing) step function mapping raw scores to calibrated probabilities — it assumes only that a higher score should mean a higher probability, nothing about the shape.

Its strength#

Because it is model-free, it can correct arbitrary monotonic miscalibration that parametric methods (Platt, temperature) miss — and with enough calibration data it typically outperforms them. It is also a general tool for monotonic regression, not only calibration.

Its weakness#

That flexibility makes it prone to overfitting when calibration data is scarce, and the piecewise-constant fit is less smooth. Unlike temperature scaling it does not guarantee the model’s accuracy is preserved.

Theme: Probability Calibration · All terminology

Hint

Mind map — connected ideas

Platt Scaling · Temperature Scaling · Confidence Level · Overconfident · Regression Models · Underconfident

Hint

More in Probability Calibration

Adaptive ECE (Expected Calibration Error with Adaptive Binning) · Brier Score · Calibration quality (Model Calibration) · Expected Calibration Error (ECE) · Maximum Calibration Error (MCE) · Murphy’s Decomposition · Overconfident · Platt Scaling · Reliability Curves (also called Calibration Curves) · Temperature Scaling · Underconfident