Adjusting Regression Models for Conditional Uncertainty Calibration
This work addresses conditional uncertainty calibration in regression models, which is important for high-stakes decision-making, but it appears incremental as it builds on existing split conformal prediction methods.
The paper tackles the problem of conditional coverage guarantees in conformal prediction for high-stakes decisions by proposing a novel algorithm to train regression functions, establishing an upper bound for miscoverage gaps and demonstrating empirical efficacy on synthetic and real-world datasets.
Conformal Prediction methods have finite-sample distribution-free marginal coverage guarantees. However, they generally do not offer conditional coverage guarantees, which can be important for high-stakes decisions. In this paper, we propose a novel algorithm to train a regression function to improve the conditional coverage after applying the split conformal prediction procedure. We establish an upper bound for the miscoverage gap between the conditional coverage and the nominal coverage rate and propose an end-to-end algorithm to control this upper bound. We demonstrate the efficacy of our method empirically on synthetic and real-world datasets.