Optical QAM Neural Networks for Efficient AI Accelerators
Analog optical neural networks are a promising paradigm for low power AI accelerators, but analog modulation even at low bit precision is energy intensive. Inspired by telecom’s QAM, QAM ONNs leverage light’s complex amplitude to boost accuracy and efficiency, creating new opportunities for hardware constrained AI applications.
Authors
- Dirk Englund (ORCID: https://orcid.org/0000-0002-1043-3489)
- Sri Krishna Vadlamani (ORCID: https://orcid.org/0000-0003-4602-0771)
- Kfir Sulimany
- Marc Gong Bacvanski
Publication Details
- Published
- 2025-01-01
- DOI
- https://doi.org/10.1364/cleo_at.2025.jss129_1
- Primary Topic
- Analog and Mixed-Signal Circuit Design
- Type
- article
- Field-Weighted Citation Impact
- 0.00