Energy-Efficient Photonic Spiking Neural Network on a monolithic silicon CMOS photonic platform
- Resource Type
- Authors
- S. J. Ben Yoo; Yun-Jhu Lee; Xian Xiao; Mehmet Berkay On
- Source
- OFC
- Subject
- Spiking neural network
Silicon
Artificial neural network
business.industry
Computer science
chemistry.chemical_element
Hardware_PERFORMANCEANDRELIABILITY
CMOS
chemistry
Electronic engineering
Spike (software development)
Photonics
business
MNIST database
Efficient energy use
- Language
We designed, simulated, and taped-out a photonic spiking neural network on a monolithic silicon CMOS photonic platform. Benchmarking shows proposed PSNN outperforms other neuromorphic hardware with 21.09fJ/spike and 61.4 W average power at MNIST experiment.