Photonic KAN: a Kolmogorov-Arnold Network Inspired Efficient Neuromorphic Accelerator
We propose Photonic Kolmogorov-Arnold Networks, leveraging optical non-linear transfer functions along edges. It achieves 2300 × reduction in footprint-energy efficiency, alongside a 7 × reduction in latency compared to previous photonic accelerators.
Authors
- Stanley Cheung (ORCID: https://orcid.org/0000-0002-4886-0013)
- Thomas Van Vaerenbergh (ORCID: https://orcid.org/0000-0002-7301-8610)
- Yiwei Peng (ORCID: https://orcid.org/0000-0002-5014-8527)
- Marco Fiorentino (ORCID: https://orcid.org/0000-0001-8867-3098)
- Raymond G. Beausoleil (ORCID: https://orcid.org/0000-0003-1139-1660)
- Xian Xiao (ORCID: https://orcid.org/0000-0002-1415-839X)
- Bassem Tossoun (ORCID: https://orcid.org/0000-0003-3608-8369)
- Sean Hooten (ORCID: https://orcid.org/0000-0003-1260-412X)
- Yuan Yuan (ORCID: https://orcid.org/0000-0001-7964-6514)
- Xinling Yu
Institutions
- Hewlett Packard Enterprise (United States) (US)
Publication Details
- Published
- 2025-01-01
- DOI
- https://doi.org/10.1364/ofc.2025.w2a.40
- Citations
- 2
- Primary Topic
- Neural Networks and Reservoir Computing
- Type
- article
- Field-Weighted Citation Impact
- 4.96