Related Experiment Video
Updated: Mar 19, 2026
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
1.2K
Image classification using optical spiking neural networks based on VCSEL with saturable absorption region neurons
Applied Optics
|March 17, 2026
Summary
This study introduces a novel neuromorphic optical framework using vertical cavity surface-emitting lasers with saturable absorbers (VCSEL-SA) for energy-efficient image classification. The proposed optical spiking neural network (OSNN) achieves 92.9% accuracy on the CIFAR-10 dataset.
Area of Science:
- Optoelectronics
- Neuromorphic Computing
- Artificial Intelligence
Background:
- Neuromorphic computing systems mimic the brain for energy efficiency.
- Vertical cavity surface-emitting lasers with saturable absorbers (VCSEL-SA) offer potential for optical neural networks.
Purpose of the Study:
- To propose and validate a neuromorphic optical framework for image classification using VCSEL-SA devices.
- To implement an optical spiking neural network (OSNN) based on ResNet-20 architecture.
Main Methods:
- A physics-based model simulated optical spiking dynamics in VCSEL-SA neurons.
- Input images were converted to spike trains using latency coding.
- VCSEL-SA layers replaced ReLU activations in a ResNet-20 CNN, trained with surrogate gradient descent.
Main Results:
- The proposed VCSEL-SA OSNN achieved 92.9% classification accuracy on the CIFAR-10 dataset.
- Spiking dynamics were encoded via temporal correlation of laser spikes.
- Hierarchical classification was performed through calibrated photonic dynamics.
Conclusions:
- The VCSEL-SA OSNN demonstrates feasibility for high-precision visual processing.
- This framework offers a promising direction for energy-efficient neuromorphic computing.
