Using unsupervised learning to detect broken symmetries, with relevance to searches for parity violation in nature.

Published: 20 Oct 2022, Last Modified: 28 Feb 2023Accepted by TMLREveryoneRevisionsBibTeX
Abstract: Testing whether data breaks symmetries of interest can be important to many fields. This paper describes a simple way that machine learning algorithms (whose outputs have been appropriately symmetrised) can be used to detect symmetry breaking. The original motivation for the paper was an important question in Particle Physics: "Is parity violated at the LHC in some way that no-one has anticipated?" and so we illustrate the main idea with an example strongly related to that question. However, in order that the key ideas be accessible to readers who are not particle physicists but who are interesting in symmetry breaking, we choose to illustrate the method/approach with a 'toy' example which places a simple discrete source of symmetry breaking (the handedness of human handwriting) within a idealised particle-physics-like context. Readers interested in seeing extensions to continuous symmetries, non-ideal environments or more realistic particle-physics contexts are provided with links to separate papers which delve into such details.
Submission Length: Regular submission (no more than 12 pages of main content)
Changes Since Last Submission: The text of this revision is (or should be) the same as the last one except for \usepackage{tmlr} has been changed to \usepackage[preprint]{tmlr} which is our (perhaps incorrect) assumption of what we are now supposed to use. The supplementary information from the last upload has now also been deleted, since that previous supplementary information was just referee responses which are not intended to be published with the accepted paper.
Assigned Action Editor: ~Jean_Barbier2
License: Creative Commons Attribution 4.0 International (CC BY 4.0)
Submission Number: 369