Ultrafast neuromorphic photonic image processing with a VCSEL neuron

Abstract

The ever-increasing demand for artificial intelligence (AI) systems is underlining a significant requirement for new, AI-optimised hardware. Neuromorphic (brain-like) processors are one highly-promising solution, with photonic-enabled realizations receiving increasing attention. Among these, approaches based upon vertical cavity surface emitting lasers (VCSELs) are attracting interest given their favourable attributes and mature technology. Here, we demonstrate a hardware-friendly neuromorphic photonic spike processor, using a single VCSEL, for all-optical image edge-feature detection. This exploits the ability of a VCSEL-based photonic neuron to integrate temporally-encoded pixel data at high speed; and fire fast (100 ps-long) optical spikes upon detecting desired image features. Furthermore, the photonic system is combined with a software-implemented spiking neural network yielding a full platform for complex image classification tasks. This work therefore highlights the potential of VCSEL-based platforms for novel, ultrafast, all-optical neuromorphic processors interfacing with current computation and communication systems for use in future light-enabled AI and computer vision functionalities.

GHz Rate Neuromorphic Photonic Spiking Neural Network With a Single Vertical-Cavity Surface-Emitting Laser (VCSEL)

“…1 (b)). The generated optical spike trains were then directly fed into the software-implemented SNN which provided a classification of the hand-written digits, achieving an average classification accuracy of 96.1% [27]. These demonstrations therefore highlight a few recent reports of promising spike-based image processing systems built using hardware-friendly neuromorphic VCSELs.…”

“…Therefore, these early works have since opened the way for new reports of spiking VCSEL neurons capable of brain-inspired processing tasks (e.g. temporal pattern recognition and image edge feature detection) [23], [24], [25], [26], [27], [28]. The first experimental processing task performed by a spiking VCSEL neuron was the coincidence detection of two input pulses and the subsequent recognition of 4-bit input patterns [23].…”

“…1 (c)) to demonstrate the potential for spike-based image encoding and processing. Finally, edge-feature detection and classification of MNIST handwritten digits was recently reported with a single experimental spiking VCSEL neuron working together with a software-implemented SNN [27]. In this demonstration a total of 5000 images were processed using the VCSEL neuron and 6 symmetrical kernels, identifying edge features with fast spiking responses (Fig.…”

“…It was also found that the intrinsic polarization properties of VCSELs provided additional avenues to enhanced performance via the control of light polarization in both injection The spike rate encoding and decoding of image colour channels and the final spike rate reconstruction. a) -c) Reprinted with permission from [23], [27], [28], under the terms of the Creative Commons CC BY license. and feedback channels.…”

Photonic Multiplexing Techniques for Optical Neuromorphic Computing

“…Since then, various lasers have been exploited as excitable devices for spiking neurons, such as distributed feedback lasers [142][143][144][145], VCSELs [146][147][148][149][150][151], and excitable fiber lasers [152][153][154][155]. For instance, M. Nahmias et al proposed a spiking neuron based on photodetectors and DFB lasers (with saturable absorbers inside), in which the input spiking signals were weighted by a tunable filter and summed by the photodetector, then drove the excitable laser to output spikes [142].…”