VCSEL Optical Neural Networks for High-throughput AI Training
We demonstrate large-scale VCSEL-based computing for convolution neural networks with spatial fanout and high clockrates. We achieved 9 parallel channels with >5 bits precision at 100-million shot-per-second, with in-system training over 93.4% accuracy.
Authors
- Niels Heermeier (ORCID: https://orcid.org/0000-0001-9340-8264)
- James A. Lott (ORCID: https://orcid.org/0000-0003-4094-499X)
- Ran Yin (ORCID: https://orcid.org/0000-0002-0056-8347)
- Mengjie Yu (ORCID: https://orcid.org/0000-0002-7815-4195)
- Yuanhao Liang
- Stephan Reitzenstein (ORCID: https://orcid.org/0000-0002-1381-9838)
- Zaijun Chen (ORCID: https://orcid.org/0000-0003-1069-3122)
- Shaoyuan Ou
- James Z. Wang (ORCID: https://orcid.org/0000-0003-4379-4173)
- Ke Xue (ORCID: https://orcid.org/0000-0001-6789-2670)
- Yuan Li (ORCID: https://orcid.org/0000-0001-5424-1764)
- Lian Zhou (ORCID: https://orcid.org/0000-0002-7705-5945)
- Tobias Heuser
- Xinyi Ren
Publication Details
- Published
- 2025-01-01
- DOI
- https://doi.org/10.1364/cleo_si.2025.ss126_2
- Primary Topic
- Neural Networks and Reservoir Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00