Reconfigurable nonlinear optical computing device for retina-inspired computing

TL;DR

This paper introduces a reconfigurable optical nonlinear device using VCSELs for retina-inspired computing, demonstrating low-power operation, programmability, and improved neural network adaptability with high recognition accuracy.

Contribution

It presents a novel sigmoid-type nonlinear optical device based on VCSELs that is reconfigurable and operates at low power, enhancing optical neural network performance.

Findings

Achieved 97.3% accuracy in handwriting recognition.

Demonstrated significant accuracy improvements under noisy and low-contrast conditions.

Showed low-power operation at 3-250 μW input power.

Abstract

Optical neural networks are at the forefront of computational innovation, utilizing photons as the primary carriers of information and employing optical components for computation. However, the fundamental nonlinear optical device in the neural networks is barely satisfied because of its high energy threshold and poor reconfigurability. This paper proposes and demonstrates an optical sigmoid-type nonlinear computation mode of Vertical-Cavity Surface-Emitting Lasers (VCSELs) biased beneath the threshold. The device is programmable by simply adjusting the injection current. The device exhibits sigmoid-type nonlinear performance at a low input optical power ranging from merely 3-250 {\mu}W. The tuning sensitivity of the device to the programming current density can be as large as 15 {\mu}W*mm2/mA. Deep neural network architecture based on such device has been proposed and demonstrated by simulation on recognizing hand-writing digital dataset, and a 97.3% accuracy has been achieved. A step further, the nonlinear reconfigurability is found to be highly useful to enhance the adaptability of the networks, which is demonstrated by significantly improving the recognition accuracy by 41.76%, 19.2%, and 25.89% of low-contrast hand-writing digital images under high exposure, low exposure, and high random noise respectively.