Related Experiment Video
Updated: Sep 17, 2025
05:57
Characterization of SiN Integrated Optical Phased Arrays on a Wafer-Scale Test Station
Published on: April 1, 2020
8.1K
All-optical convolutional neural network based on phase change materials in silicon photonics platform.
1Department of Communications and Electronics, School of Electrical and Computer Engineering, Shiraz University, Shiraz, Iran.
Scientific Reports
|July 2, 2025
Summary
This study introduces an all-optical convolution neural network using silicon photonics for efficient AI. The novel design simplifies complexity and achieves high accuracy in data classification tasks.
Area of Science:
- Photonics
- Artificial Intelligence
- Computer Engineering
Background:
- Current neural networks often rely on complex electronic components.
- Integrated optical elements offer potential for reduced power consumption and increased speed.
- Silicon photonics provides a mature platform for optical device fabrication.
Purpose of the Study:
- To design and simulate an integrated all-optical convolution neural network (CNN).
- To implement convolution, max-pooling, and fully connected layers using silicon photonics.
- To reduce overall network complexity and reliance on electro-optical elements.
Main Methods:
- Utilized GST-based active waveguides for network layers.
- Mitigated ReLU requirement in convolution layers using positive kernel values.
- Employed finite-difference time-domain (FDTD) method and coupled mode theory for simulations.
- Validated network performance using Python programming.
Main Results:
- Achieved 91.90% accuracy in MNIST data classification.
- Demonstrated 80% accuracy in signal modulation identification (RML2016.10a dataset).
- The all-optical CNN design shows comparable performance to electronic counterparts with reduced complexity.
Conclusions:
- The proposed integrated all-optical CNN offers a simplified and efficient design.
- Silicon photonics platform enables a fully optical implementation of complex neural network layers.
- This approach paves the way for high-performance, low-complexity optical computing.
