Efficient machine unlearning with minimax optimality

Jingyi Xie, Linjun Zhang · 2026

Read Efficient machine unlearning with minimax optimality by Jingyi Xie, Linjun Zhang (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-2604-05669.txt.

Description

Xie, Zhang, and Li frame machine unlearning statistically, seeking removal of data influence without full retraining. For squared loss their Unlearning Least Squares method proves minimax optimal using only the trained estimator, forget samples, and a small retained subsample. Error splits into an oracle term plus an unlearning cost, and valid inference follows without retraining.

Catalog record

CREATORJingyi Xie, Linjun Zhang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE171881
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.05669

Browse the full library · About the Tower of Babel