Minimum Volume Conformal Sets for Multivariate Regression
This work addresses the need for efficient and accurate predictive uncertainty quantification in multivariate regression, offering a novel method that improves upon existing approaches with rigid assumptions or high computational costs.
The paper tackles the problem of constructing predictive sets for multivariate regression with finite-sample validity, proposing an optimization-driven framework that directly learns minimum-volume covering sets while ensuring valid coverage, achieving tight and computationally efficient prediction sets as demonstrated on real-world datasets.
Conformal prediction provides a principled framework for constructing predictive sets with finite-sample validity. While much of the focus has been on univariate response variables, existing multivariate methods either impose rigid geometric assumptions or rely on flexible but computationally expensive approaches that do not explicitly optimize prediction set volume. We propose an optimization-driven framework based on a novel loss function that directly learns minimum-volume covering sets while ensuring valid coverage. This formulation naturally induces a new nonconformity score for conformal prediction, which adapts to the residual distribution and covariates. Our approach optimizes over prediction sets defined by arbitrary norm balls, including single and multi-norm formulations. Additionally, by jointly optimizing both the predictive model and predictive uncertainty, we obtain prediction sets that are tight, informative, and computationally efficient, as demonstrated in our experiments on real-world datasets.