Learning Broken Symmetries with Approximate Invariance

December 25, 2024 Β· Declared Dead Β· πŸ› Physical Review D

πŸ‘» CAUSE OF DEATH: Ghosted
No code link whatsoever

"No code URL or promise found in abstract"

Evidence collected by the PWNC Scanner

Authors Seth Nabat, Aishik Ghosh, Edmund Witkowski, Gregor Kasieczka, Daniel Whiteson arXiv ID 2412.18773 Category hep-ph Cross-listed cs.LG, hep-ex Citations 3 Venue Physical Review D Last Checked 3 months ago
Abstract
Recognizing symmetries in data allows for significant boosts in neural network training, which is especially important where training data are limited. In many cases, however, the exact underlying symmetry is present only in an idealized dataset, and is broken in actual data, due to asymmetries in the detector, or varying response resolution as a function of particle momentum. Standard approaches, such as data augmentation or equivariant networks fail to represent the nature of the full, broken symmetry, effectively overconstraining the response of the neural network. We propose a learning model which balances the generality and asymptotic performance of unconstrained networks with the rapid learning of constrained networks. This is achieved through a dual-subnet structure, where one network is constrained by the symmetry and the other is not, along with a learned symmetry factor. In a simplified toy example that demonstrates violation of Lorentz invariance, our model learns as rapidly as symmetry-constrained networks but escapes its performance limitations.
Community shame:
Not yet rated
Community Contributions

Found the code? Know the venue? Think something is wrong? Let us know!

πŸ“œ Similar Papers

In the same crypt β€” hep-ph

Died the same way β€” πŸ‘» Ghosted