Running summarizing tools on a new article

Check the paper below, then start its synthesis.


A method to challenge symmetries in data with self-supervised learning

Rupert Tombs, Christopher G. Lester

License: CC BY 4.0

Abstract: Symmetries are key properties of physical models and experimental designs, but may or may not be realized in nature. We propose a practical and general method to test such suspect symmetries in data. Self-supervision, which derives learning objectives from data, is used to train models to predict 'which is real?' between real and symmetrically transformed data; if a model predicts successfully in independent tests, then its target symmetry is challenged. Crucially, our method handles filtered data, which often arise from inefficiencies or deliberate selections, and which can give the illusion of asymmetry if mistreated. Examples are used to demonstrate how the method works and how the models' predictions can be interpreted.

Submitted to arXiv on 09 Nov. 2021