Geometry-Aware Uncertainty Quantification via Conformal Prediction on Manifolds

Marzieh Amiri Shahbazi, Ali Baheri · 2026

Read Geometry-Aware Uncertainty Quantification via Conformal Prediction on Manifolds by Marzieh Amiri Shahbazi, Ali Baheri (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-16015.txt.

Description

Amiri Shahbazi and Baheri adapt conformal prediction to answers that live on curved spaces. Standard conformal methods guarantee coverage for regression, but they assume flat output spaces, so their regions can mislead when responses sit on a manifold such as a sphere.

Their adaptive geodesic method scores errors by geodesic distance and scales those scores by a cross validated estimate of local difficulty, handling noise that varies across the space. On the sphere it yields caps of position independent area whose radii still adapt to local noise. Tests on a synthetic sphere task and on geomagnetic field forecasting keep valid overall coverage, steady conditional coverage, and stronger worst case coverage than flat or non adaptive baselines.

Catalog record

CREATORMarzieh Amiri Shahbazi, Ali Baheri
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE26 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-16015

Browse the full library · About the Tower of Babel