Lorentz Local Canonicalization: How to Make Any Network Lorentz-Equivariant

Jonas Spinner, Luigi Favaro, Peter Lippmann, Sebastian Pitz, Gerrit Gerhartz, Tilman Plehn, Fred A. Hamprecht · 2025

Read Lorentz Local Canonicalization: How to Make Any Network Lorentz-Equivariant by Jonas Spinner, Luigi Favaro, Peter Lippmann, Sebastian Pitz, Gerrit Gerhartz, Tilman Plehn, Fred A. Hamprecht (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-2505-20280.txt.

Description

Neural networks for particle physics must respect relativity: shuffle the reference frame and the physics should not change. Building that symmetry in usually means bespoke layers. Jonas Spinner and co-authors offer a universal adapter in this 2025 paper, Lorentz Local Canonicalization. The network predicts a local reference frame for each particle, re-expresses everything in that frame, and lets any ordinary architecture do the rest.

The result is exactly Lorentz-equivariant versions of transformers and graph networks, with data augmentation falling out as a special case. On standard particle physics tasks the models matched or beat specialized rivals while running four times faster on a tenth of the computing. The authors pitch the trick as a general recipe.

Catalog record

CREATORJonas Spinner, Luigi Favaro, Peter Lippmann, Sebastian Pitz, Gerrit Gerhartz, Tilman Plehn, Fred A. Hamprecht
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE79 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-20280

Browse the full library · About the Tower of Babel