Assessment of Prediction Intervals Using Uncertainty Characteristics Curves

TL;DR

This paper introduces a new method using Uncertainty Characteristics Curves to evaluate prediction intervals in regression, providing a comprehensive, operating point agnostic assessment of model uncertainty.

Contribution

It proposes a novel assessment methodology based on operating characteristics curves and gain over a null reference, enhancing evaluation of prediction intervals.

Findings

Demonstrates utility of Uncertainty Characteristics Curves in selected scenarios

Addresses the need for comprehensive prediction interval assessment

Provides a new tool for uncertainty quantification in AI models

Abstract

Accurate quantification of model uncertainty has long been recognized as a fundamental requirement for trusted AI. In regression tasks, uncertainty is typically quantified using prediction intervals calibrated to an ad-hoc operating point, making evaluation and comparison across different studies relatively difficult. Our work leverages: (1) the concept of operating characteristics curves and (2) the notion of a gain over a null reference, to derive a novel operating point agnostic assessment methodology for prediction intervals. The paper defines the Uncertainty Characteristics Curve and demonstrates its utility in selected scenarios. We argue that the proposed method addresses the current need for comprehensive assessment of prediction intervals and thus represents a valuable addition to the uncertainty quantification toolbox.