Photonic Unitary Neural Network Chip with Complex-valued Transformation Layer
- Resource Type
- Conference
- Authors
- Wang, Ruiting; Wang, Pengfei; Luo, Guangzhen; Yu, Hongyan; Zhou, Xuliang; Zhang, Yejin; Wang, Wei; Pan, Jiaoqing
- Source
- 2021 Asia Communications and Photonics Conference (ACP) Communications and Photonics Conference (ACP), 2021 Asia. :1-3 Oct, 2021
- Subject
- Communication, Networking and Broadcast Technologies
Components, Circuits, Devices and Systems
Fields, Waves and Electromagnetics
Photonics and Electrooptics
Power, Energy and Industry Applications
Signal Processing and Analysis
Fourier transforms
Neural networks
Asia
Photonics
(250.5300) Photonic integrated circuits
(200.4260) Neural networks
- Language
We propose a complex-valued transformation layer whose architecture is compatible with the photonic unitary neural network chip. We simulate unitary neural networks on the MNIST dataset and complex-valued transformation layer can significantly improve the accuracy.