Gradient Descent Optimization of Diffractive Neural Networks for High-capacity, Scalable and Low Crosstalk Mode-Wavelength Multiplexing
We demonstrate a gradient descent-based optimization framework for designing the high-capacity, scalable, and low crosstalk mode-wavelength multiplexer using diffractive neural networks, which can effectively reduce the complexity and volume of high-speed optical communication multiplexing systems.
Authors
- Shaoxiang Duan (ORCID: https://orcid.org/0000-0002-6583-7675)
- Zehui Lu (ORCID: https://orcid.org/0000-0002-7312-7075)
- Hao Zhang (ORCID: https://orcid.org/0000-0001-9618-8746)
- Wei Lin (ORCID: https://orcid.org/0000-0001-9711-2022)
- Haifeng Liu
- Lei Zhu (ORCID: https://orcid.org/0000-0003-1685-7968)
Publication Details
- Published
- 2025-01-01
- DOI
- https://doi.org/10.1364/cleo_si.2025.ss132_3
- Primary Topic
- Orbital Angular Momentum in Optics
- Type
- article
- Field-Weighted Citation Impact
- 0.00