conference-paper
Tunable Nonlinear Activation Functions for Optical Neural Networks
Research footprint
At a glance
- Citations
- 5
- References
- 4
- Comments
- 0
Paper overview
Abstract
We introduce an electro-optic hardware platform for realizing optical nonlin earities and demonstrate that, as a nonlinear activation function, it can substantially improve the classification performance of optical neural networks.
Record transparency
Publication details
- DOI
- 10.1364/cleo_si.2020.sm1e.2
- OpenAlex
- W3082288064
- Document type
- conference-paper
- Language
- EN
- Source
- Conference on Lasers and Electro-Optics
- Last metadata update
Comments
Log in to join the discussion.