Variable-Based Calibration for Machine Learning Classifiers

Markelle Kelly, Padhraic Smyth

The deployment of machine learning classifiers in high-stakes domains requires well-calibrated confidence scores for model predictions. In this paper we introduce the notion of var…