Abstract:The computational demands of modern AI have spurred interest in optical neural networks (ONNs) which offer the potential benefits of increased speed and lower power consumption. However, current ONNs face various challenges,most significantly a limited calculation precision (typically around 4 bits) and the requirement for high-resolution signal format converters (digital-to-analogue conversions (DACs) and analogue-to-digital conversions (ADCs)). These challenges are inherent to their analog computing nature and pose significant obstacles in practical implementation. Here, we propose a digital-analog hybrid optical computing architecture for ONNs, which utilizes digital optical inputs in the form of binary words. By introducing the logic levels and decisions based on thresholding, the calculation precision can be significantly enhanced. The DACs for input data can be removed and the resolution of the ADCs can be greatly reduced. This can increase the operating speed at a high calculation precision and facilitate the compatibility with microelectronics. To validate our approach, we have fabricated a proof-of-concept photonic chip and built up a hybrid optical processor (HOP) system for neural network applications. We have demonstrated an unprecedented 16-bit calculation precision for high-definition image processing, with a pixel error rate (PER) as low as $1.8\times10^{-3}$ at an signal-to-noise ratio (SNR) of 18.2 dB. We have also implemented a convolutional neural network for handwritten digit recognition that shows the same accuracy as the one achieved by a desktop computer. The concept of the digital-analog hybrid optical computing architecture offers a methodology that could potentially be applied to various ONN implementations and may intrigue new research into efficient and accurate domain-specific optical computing architectures for neural networks.

Photonic single perceptron at Giga-OP/s speeds with Kerr microcombs for scalable optical neural networks

Single photonic perceptron based on a soliton crystal Kerr microcomb for high-speed, scalable, optical neural networks

Optical neuromorphic processing at Tera-OP/s speeds based on Kerr soliton crystal microcombs

Optical neuromorphic processing with Kerr microcombs: Scaling the network in size and speed to the PetaOp regime

11 Tera-OPs/s Photonic Convolutional Accelerator and Deep Optical Neural Network Based on an Integrated Kerr Soliton Crystal Microcomb

A photonic complex perceptron for ultrafast data processing

Optical neuromorphic processing with soliton crystal Kerr microcombs: Scaling the network in size and speed

11 TOPS photonic convolutional accelerator for optical neural networks

A complete, parallel and autonomous photonic neural network in a semiconductor multimode laser

Neuromorphic computing using wavelength-division multiplexing

Neuromorphic Computing Based on Wavelength-Division Multiplexing

Electrically Programmable Phase-Change Photonic Memory for Optical Neural Networks with Nanoseconds in Situ Training Capability

InP Photonic Integrated Multi-Layer Neural Networks: Architecture and Performance Analysis

Digital-analog hybrid matrix multiplication processor for optical neural networks

11 TeraFLOPs per second photonic convolutional accelerator for deep learning optical neural networks

Optical Implementation of Neural Learning Algorithms Based on Cross-gain Modulation in a Semiconductor Optical Amplifier

Deep learning with coherent VCSEL neural networks

Integrated multi-operand optical neurons for scalable and hardware-efficient deep learning

Single-shot optical neural network

Photonic Neural Networks and Its Applications