Symmetry-Informed Deep Learning for Scattering Parameter Prediction
Paper i proceeding, 2026
Deep learning surrogate models can accelerate modelling of electromagnetic devices by several orders of magnitude, but usually require very large amounts of training data. We show that neural networks can be made more data efficient by utilizing symmetries of Maxwell’s equations.
machine learning
diffraction gratings
metasurfaces
symmetries
scattering
inverse design
Författare
Viktor Aadland Lilja
Chalmers, Fysik, Kondenserad materie- och materialteori
Philippe Tassin
Chalmers, Fysik, Kondenserad materie- och materialteori
International Conference on Metamaterials, Photonic Crystals and Plasmonics
24291390 (eISSN)
1119-1120
16th International Conference on Metamaterials, Photonic Crystals and Plasmonics, META 2026
Dublin, Ireland,
Dublin, Ireland,
Ämneskategorier (SSIF 2025)
Datavetenskap (datalogi)