Photonic Perceptron Based on a Kerr Microcomb for High‐Speed, Scalable, Optical Neural Networks

Abstract

Optical artificial neural networks (ONNs)-analog computing hardware tailored for machine learning-have significant potential for achieving ultrahigh computing speed and energy efficiency. A new approach to architectures for ONNs based on integrated Kerr microcomb sources that is programmable, highly scalable, and capable of reaching ultra-high speeds is proposed here. The building block of the ONN-a single neuron perceptron-is experimentally demonstrated that reaches a high single-unit throughput speed of 11.9 Giga-FLOPS at 8 bits per FLOP, corresponding to 95.2 Gbps, achieved by mapping synapses onto 49 wavelengths of a microcomb. The perceptron is tested on simple standard benchmark datasets-handwritten-digit recognition and cancer-cell detection-achieving over 90% and 85% accuracy, respectively. This performance is a direct result of the record low wavelength spacing (49 GHz) for a coherent integrated microcomb source, which results in an unprecedented number of wavelengths for neuromorphic optics. Finally, an approach to scaling the perceptron to a deep learning network is proposed using the same single microcomb device and standard off-the-shelf telecommunications technology, for high-throughput operation involving full matrix multiplication for applications such as real-time massive data processing for unmanned vehicles and aircraft tracking.

Photonic RF fractional Hilbert transformers and spectral filters based on 49GHz FSR soliton crystal microcombs

“…We have demonstrated photonic RF signal processors based on soliton crystal microcombs. This approach offers a greatly reduced footprint, lower complexity, and potentially lower cost, representing a solid step towards the miniaturization of high performance photonic RF systems [16][17][18][19][20][21][22][23][24][25].…”

RF and microwave photonic high bandwidth signal processing based on Kerr micro-comb sources

“…Many cutting-RF and microwave photonic high bandwidth signal processing based on Kerr micro-comb sources Mengxi Tan, 1 Xingyuan Xu, 1,* Jiayang Wu, 1 Roberto Morandotti, 2 Arnan Mitchell, 3 and David J. Moss 1 edge applications have been demonstrated based on CMOS-compatible micro-combs, ranging from filter-driven modelocked lasers [49][50][51][52] to quantum physics [53][54][55][56][57][58][59][60][61]. Most recently, Kerr micro-combs have demonstrated their enormous potential as sources for ultrahigh bandwidth coherent optical fiber communications [62][63][64], optical neural networks [65][66][67], optical frequency synthesis [68], and have been reviewed extensively [47, 69 -76]. The success of these new CMOS platforms as well as other CMOS platforms such as amorphous silicon [77] and silicon rich silicon nitride (SRSN) [78] arises from a combination of their low linear loss, moderate to high nonlinearity and low or even negligible nonlinear loss (TPA).…”

Orthogonally polarized RF single sideband generation with Kerr microcombs

“…Its main applications lie in the areas of radar, communications and sensing. Many different capabilities have been achieved, such as advanced encoding format modulators [5][6][7][8][9][10], devices for RF beam shaping [11], RF signal processors based on transverse filtering [12][13][14], RF channelizers [15] and spectrum analysis [16][17][18], and much more [19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38].…”