AI engineering glossary · Evaluation metrics
What is platt scaling?
Platt scaling (also called sigmoid calibration): A post-processing method that fits a logistic regression on a model's raw scores, using held-out data, to convert them into calibrated probabilities. It has just two parameters, so it works with little data. It was introduced for support vector machines.
Platt scaling: an example
An SVM outputs distances from the decision boundary; Platt scaling learns a sigmoid mapping those distances to probabilities such as 0.83.
Related terms
More Evaluation metrics terms
- Precision at k
- Precision-recall break-even point
- Precision-recall curve
- Precision-recall trade-off
- Prediction interval coverage
- Prevalence
- Proper scoring rule
- R-precision
Where you learn it
Platt scaling is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.