Deep learning based on the loss metric for inverse design of photonic resonators
Optics Express · 26.08.2024 · 32 (18) · 31193
DOI: 10.1364/oe.523652 ↗Kurz gefasst
Fachartikel in der Zeitschrift Optics Express, online erschienen am 13.08.2024. Verfasst von Hao You und 4 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 23 Literaturangaben. Gefördert von: National Key Research and Development Program of China.
Abstract (Originalsprache)
We have proposed and implemented what we believe to be a novel metric for providing a more effective similarity evaluation to the deep learning algorithms used for the inverse design of resonant photonic devices. The conventional loss functions, such as mean square error (MSE) and mean absolute error (MAE), are incapable of recognizing the characteristics of resonances accurately. Therefore, we have calculated the time domain complex vectors through the Fourier transform (FT) of the original desired spectra, and the complex results containing amplitude and phase could distinguish the resonances more significantly. Our new loss metric considers both the spectral MSE and the time domain vector error (TVE), and test results demonstrate that this new technique could realize a more effective resonance line shape match and a lower test error compared to the existing loss evaluation methods.
Förderung
- National Key Research and Development Program of China
Zitieren
APA
You, H., Du, L., Xu, X., Sun, C., & Zhao, J. (2024). Deep learning based on the loss metric for inverse design of photonic resonators. Optics Express, 32(18), 31193. https://doi.org/10.1364/oe.523652
Vancouver
You H, Du L, Xu X, Sun C, Zhao J. Deep learning based on the loss metric for inverse design of photonic resonators. Optics Express. 2024;32(18):31193. doi:10.1364/oe.523652
BibTeX
@article{You2024,
author = {You, Hao and Du, Liuge and Xu, Xiao and Sun, Chonglei and Zhao, Jia},
title = {Deep learning based on the loss metric for inverse design of photonic resonators},
journal = {Optics Express},
year = {2024},
volume = {32},
number = {18},
pages = {31193},
doi = {10.1364/oe.523652}
}
EUVIMED