Low-distortion and GPU-compatible Tree Embeddings in Hyperbolic Space

Max van Spengler, Pascal Mettes · 2025

Read Low-distortion and GPU-compatible Tree Embeddings in Hyperbolic Space by Max van Spengler, Pascal Mettes (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-2502-17130.txt.

Description

Max van Spengler and Pascal Mettes improve how tree-shaped data, like family trees of concepts, are embedded in curved hyperbolic space. Their 2025 paper starts from a known strength: curved space fits trees far better than flat space. Two flaws held back the best building method, poor spacing of anchor points and a need for extra-precise arithmetic that GPUs cannot run.

Their fix weaves angular spacing into the construction, spreading anchors by a Delaunay-style rule to cut distortion. They also rebuild the math on floating-point expansion arithmetic, which keeps quality high while running on graphics hardware. The result is tree embeddings that stay faithful and can sit inside deep learning models. The authors aim at hierarchies, taxonomies, and knowledge bases.

Catalog record

CREATORMax van Spengler, Pascal Mettes
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE73 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-17130

Browse the full library · About the Tower of Babel