Distance-Based Tree-Sliced Wasserstein Distance
Hoang V. Tran ∗, Khoi N.M. Nguyen ∗, [email protected], Trang Pham, Thanh T. Chu, National University of Singapore, Tam Le †, Tan M. Nguyen † · 2025
Read Distance-Based Tree-Sliced Wasserstein Distance by Hoang V. Tran ∗, Khoi N.M. Nguyen ∗, [email protected], Trang Pham, Thanh T. Chu, National University of Singapore, Tam Le †, Tan M. Nguyen † (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-2503-11050.txt.
Description
Hoang Tran and co-authors add a new variant in 2025 to a family of fast methods for comparing two piles of data. The exact comparison is costly, so fast methods slice the data onto lines and compare the slices. That slicing can lose facts about the shape of the data, which hurts later tasks.
A newer method slices onto tree shaped systems instead of lines, which keeps more shape facts at a like cost. The team finds a weak point in how that method splits data onto the trees. Their distance based variant aims to fix the split step while keeping the speed. Tests in the paper weigh the new variant against past ones. Shape aware comparison at low cost serves machine learning tasks from image match to model checks, where data piles are large and complex.
Catalog record
| CREATOR | Hoang V. Tran ∗, Khoi N.M. Nguyen ∗, [email protected], Trang Pham, Thanh T. Chu, National University of Singapore, Tam Le †, Tan M. Nguyen † |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 91 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-11050