Free-Space Optical Spiking Neural Network

TL;DR

The paper tackles the speed and energy bottlenecks of electronic neural processors by introducing OSCNN, an optical free-space deep spiking convolutional network inspired by the human eye. It integrates Gabor-filter-based feature extraction, intensity-to-phase latency encoding via an SLM, and an optical synchronizer, assembled into a three-layer compute path with a saturable-absorber classifier. Empirical results on MNIST, ETH-80, and Caltech show competitive accuracy (e.g., MNIST up to $95.2\%$) and favorable latency estimates (about $2.44\ \mathrm{ms}$) with emphasis on low-power operation, while experiments compare fixed versus trainable Gabor filters and analyze robustness to noise. The work points to practical optical neuromorphic designs and sets the stage for future neuron models and metasurface-enabled convolution networks to further reduce complexity and energy while maintaining high throughput.

Abstract

Neuromorphic engineering has emerged as a promising avenue for developing brain-inspired computational systems. However, conventional electronic AI-based processors often encounter challenges related to processing speed and thermal dissipation. As an alternative, optical implementations of such processors have been proposed, capitalizing on the intrinsic information-processing capabilities of light. Within the realm of optical neuromorphic engineering, various optical neural networks (ONNs) have been explored. Among these, Spiking Neural Networks (SNNs) have exhibited notable success in emulating the computational principles of the human brain. Nevertheless, the integration of optical SNN processors has presented formidable obstacles, mainly when dealing with the computational demands of large datasets. In response to these challenges, we introduce a pioneering concept: the Free-space Optical deep Spiking Convolutional Neural Network (OSCNN). This novel approach draws inspiration from computational models of the human eye. We have meticulously designed various optical components within the OSCNN to tackle object detection tasks across prominent benchmark datasets, including MNIST, ETH 80, and Caltech. Our results demonstrate promising performance with minimal latency and power consumption compared to their electronic ONN counterparts. Additionally, we conducted several pertinent simulations, such as optical intensity to-latency conversion and synchronization. Of particular significance is the evaluation of the feature extraction layer, employing a Gabor filter bank, which stands to impact the practical deployment of diverse ONN architectures significantly.

Figures (5)

• Figure 1: Outline of OSCNN.
• Figure 2: The simulation of OSCNN involves loading data from a laser and SLM and simulating different optical free space components.
• Figure 3: Gabor filters are created as sine wave filters in four different directions, namely $0$, $\frac{\pi}{2}$, $\pi$, and $\frac{3\pi}{4}$ degrees. These filters have different thicknesses and function as edge detectors, resembling the simple cells found in the primary visual cortex RN17.
• Figure 4: Accuracy of different electrical and optical Neural Network on Caltech and ETH-80 data sets.
• Figure 5: Accuracy of different electrical and optical Neural Networks for various input noise levels.