Measuring Heterogeneity in Machine Learning with Distributed Energy Distance

Mengchen Fan, Baocheng Geng, Roman Shterenberg, Joseph A. Casey, Zhong Chen, Keren Li · 2025

Read Measuring Heterogeneity in Machine Learning with Distributed Energy Distance by Mengchen Fan, Baocheng Geng, Roman Shterenberg, Joseph A. Casey, Zhong Chen, Keren Li (2025) 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-2501-16174.txt.

Description

Mengchen Fan and five co-authors measure how much data sources differ in shared machine learning, in a 2025 paper. When phones or hospitals train one model together, mismatched data slows learning and skews results. A good yardstick for mismatch would help planners react.

The authors adopt energy distance, a statistic that senses shifts between data distributions. Computing it raw is too costly at scale, so they build shortcut formulas that keep its key properties. Simulations show that tracking feature mismatch well speeds up joint training. They also propose using the measure to set penalty weights that pull stray sites' predictions into line. The tools target federated systems where data cannot be pooled for privacy reasons.

Catalog record

CREATORMengchen Fan, Baocheng Geng, Roman Shterenberg, Joseph A. Casey, Zhong Chen, Keren Li
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-16174

Browse the full library · About the Tower of Babel