Evaluating Posterior Probabilities: Decision Theory, Proper Scoring Rules, and Calibration
This work addresses a foundational problem in machine learning evaluation for researchers and practitioners, advocating for a shift from calibration metrics to proper scoring rules to assess classifier posteriors more accurately.
The paper argues that calibration metrics like expected calibration error (ECE) are inadequate for evaluating posterior probabilities in machine learning classifiers, proposing instead the use of proper scoring rules (PSRs) as a principled measure that accounts for both calibration and discrimination. It introduces a new calibration metric called calibration loss, derived from PSR decomposition, and shows it is superior to ECE and other metrics through theoretical and empirical evidence.
Most machine learning classifiers are designed to output posterior probabilities for the classes given the input sample. These probabilities may be used to make the categorical decision on the class of the sample; provided as input to a downstream system; or provided to a human for interpretation. Evaluating the quality of the posteriors generated by these system is an essential problem which was addressed decades ago with the invention of proper scoring rules (PSRs). Unfortunately, much of the recent machine learning literature uses calibration metrics -- most commonly, the expected calibration error (ECE) -- as a proxy to assess posterior performance. The problem with this approach is that calibration metrics reflect only one aspect of the quality of the posteriors, ignoring the discrimination performance. For this reason, we argue that calibration metrics should play no role in the assessment of posterior quality. Expected PSRs should instead be used for this job, preferably normalized for ease of interpretation. In this work, we first give a brief review of PSRs from a practical perspective, motivating their definition using Bayes decision theory. We discuss why expected PSRs provide a principled measure of the quality of a system's posteriors and why calibration metrics are not the right tool for this job. We argue that calibration metrics, while not useful for performance assessment, may be used as diagnostic tools during system development. With this purpose in mind, we discuss a simple and practical calibration metric, called calibration loss, derived from a decomposition of expected PSRs. We compare this metric with the ECE and with the expected score divergence calibration metric from the PSR literature and argue, using theoretical and empirical evidence, that calibration loss is superior to these two metrics.