Label Shift Estimation With Incremental Prior Update

Yunrui Zhang, Gustavo Batista, Salil S. Kanhere · 2026

Read Label Shift Estimation With Incremental Prior Update by Yunrui Zhang, Gustavo Batista, Salil S. Kanhere (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-2604-01651.txt.

Description

Zhang and colleagues propose a new approach for post-hoc label shift estimation, updating the prior on each sample and adjusting each posterior for better estimation. Label shift estimation aims to estimate the changing label distribution in the testing set, assuming the likelihood does not change. Unlike previous methods using validation-set moment matching or expectation-maximization likelihood maximization, the method relies on a weaker notion of calibration. Tests on CIFAR-10 and MNIST show it beats current state-of-the-art maximum likelihood-based methods.

Catalog record

CREATORYunrui Zhang, Gustavo Batista, Salil S. Kanhere
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44733
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.01651

Browse the full library · About the Tower of Babel