Saved in:
Bibliographic Details
Main Authors: Principato, Guillaume, Stoltz, Gilles
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.15824
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915559810007040
author Principato, Guillaume
Stoltz, Gilles
author_facet Principato, Guillaume
Stoltz, Gilles
contents We study conformal inference in non-exchangeable environments through the lens of Blackwell's theory of approachability. We first recast adaptive conformal inference (ACI, Gibbs and Candès, 2021) as a repeated two-player vector-valued finite game and characterize attainable coverage--efficiency tradeoffs. We then construct coverage and efficiency objectives under potential restrictions on the adversary's play, and design a calibration-based approachability strategy to achieve these goals. The resulting algorithm enjoys strong theoretical guarantees and provides practical insights, though its computational burden may limit deployment in practice.
format Preprint
id arxiv_https___arxiv_org_abs_2510_15824
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Blackwell's Approachability for Sequential Conformal Inference
Principato, Guillaume
Stoltz, Gilles
Machine Learning
91A26 62G08
We study conformal inference in non-exchangeable environments through the lens of Blackwell's theory of approachability. We first recast adaptive conformal inference (ACI, Gibbs and Candès, 2021) as a repeated two-player vector-valued finite game and characterize attainable coverage--efficiency tradeoffs. We then construct coverage and efficiency objectives under potential restrictions on the adversary's play, and design a calibration-based approachability strategy to achieve these goals. The resulting algorithm enjoys strong theoretical guarantees and provides practical insights, though its computational burden may limit deployment in practice.
title Blackwell's Approachability for Sequential Conformal Inference
topic Machine Learning
91A26 62G08
url https://arxiv.org/abs/2510.15824