Wavelength-Space Domain High-Throughput Artificial Neural Networks by Parallel Photoelectric Matrix Multiplier
2020
We propose a massively parallel neural network architecture with photonic matrix-vector multiplication in the wavelength and space domains with balanced photodetectors and nonlinear transfer functions in MZI modulators. An experimental proof-of-principle demonstration is also discussed.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
11
References
2
Citations
NaN
KQI