Channel Tiling for Improved Performance and Accuracy of Optical Neural Network Accelerators

TL;DR

This paper introduces a channel tiling scheme for optical CNN accelerators using 4F optics, significantly improving throughput, robustness, and reducing hardware requirements without additional optical components.

Contribution

The proposed channel tiling method enables scalable multi-channel 4F CNN systems by performing optical domain channel summation, addressing key scalability and precision challenges.

Findings

Achieves 10-50X throughput improvement

Reduces required output camera resolution by up to 3X

Improves robustness to sensing quantization and noise

Abstract

Low latency, high throughput inference on Convolution Neural Networks (CNNs) remains a challenge, especially for applications requiring large input or large kernel sizes. 4F optics provides a solution to accelerate CNNs by converting convolutions into Fourier-domain point-wise multiplications that are computationally 'free' in optical domain. However, existing 4F CNN systems suffer from the all-positive sensor readout issue which makes the implementation of a multi-channel, multi-layer CNN not scalable or even impractical. In this paper we propose a simple channel tiling scheme for 4F CNN systems that utilizes the high resolution of 4F system to perform channel summation inherently in optical domain before sensor detection, so the outputs of different channels can be correctly accumulated. Compared to state of the art, channel tiling gives similar accuracy, significantly better robustness to sensing quantization (33\% improvement in required sensing precision) error and noise (10dB reduction in tolerable sensing noise), 0.5X total filters required, 10-50X+ throughput improvement and as much as 3X reduction in required output camera resolution/bandwidth. Not requiring any additional optical hardware, the proposed channel tiling approach addresses an important throughput and precision bottleneck of high-speed, massively-parallel optical 4F computing systems.