CLEAR: Calibrated Learning for Epistemic and Aleatoric Risk
This addresses the need for balanced uncertainty estimation in regression, offering a practical calibration method for scenarios with high uncertainty, though it is incremental as it builds on existing estimators.
The paper tackles the problem of combining aleatoric and epistemic uncertainty in predictive intervals for regression tasks, proposing CLEAR which achieves an average improvement of 28.2% and 17.4% in interval width compared to baselines while maintaining coverage.
Existing methods typically address either aleatoric uncertainty due to measurement noise or epistemic uncertainty resulting from limited data, but not both in a balanced manner. We propose CLEAR, a calibration method with two distinct parameters, $γ_1$ and $γ_2$, to combine the two uncertainty components and improve the conditional coverage of predictive intervals for regression tasks. CLEAR is compatible with any pair of aleatoric and epistemic estimators; we show how it can be used with (i) quantile regression for aleatoric uncertainty and (ii) ensembles drawn from the Predictability-Computability-Stability (PCS) framework for epistemic uncertainty. Across 17 diverse real-world datasets, CLEAR achieves an average improvement of 28.2% and 17.4% in the interval width compared to the two individually calibrated baselines while maintaining nominal coverage. Similar improvements are observed when applying CLEAR to Deep Ensembles (epistemic) and Simultaneous Quantile Regression (aleatoric). The benefits are especially evident in scenarios dominated by high aleatoric or epistemic uncertainty.