Minkowski-Type Wasserstein Metrics and Barycenters for Location-Scale Mixtures with Application to Domain Adaptation

Songyan Luo, Yunxin Zhang · 2026

Read Minkowski-Type Wasserstein Metrics and Barycenters for Location-Scale Mixtures with Application to Domain Adaptation by Songyan Luo, Yunxin 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-2606-26482.txt.

Description

Songyan Luo and Yunxin Zhang build faster optimal transport methods for mixture models. Optimal transport measures how much work it takes to reshape one distribution into another. Standard discrete methods match individual sample points. Their cost grows at least with the square of the sample size. Location-scale mixtures, blends of shifted and stretched copies of one shape, offer a shortcut.

The authors define distance measures for such mixtures using generalised Minkowski inequalities. They characterise the optimal transport maps between the underlying families. A key result reduces the big matching problem to transport between mixture components only. Components are few, even when samples are many. Computing time then grows only linearly with sample size. Tests on the VisDA-C domain adaptation benchmark followed. Domain adaptation reuses a model trained in one setting in a different one. Accuracy stayed competitive while cost and memory fell substantially.

Catalog record

CREATORSongyan Luo, Yunxin Zhang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE124 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.26482

Browse the full library · About the Tower of Babel