Deep learning with coherent nanophotonic circuits

By Yichen Shen, Nicholas Christopher Harris, S. Skirlo, Mihika Prabhu, T. Baehr‐Jones, Michael Hochberg, Xin Sun, Shijie Zhao, Hugo Larochelle, Dirk Englund and 1 more. Nature Photonics, 2016

2,982 citations recorded by the source corpus.

Artificial neural networks are computational network models inspired by signal processing in the brain. These models have dramatically improved performance for many machine-learning tasks, including speech and image recognition. However, today's computing hardware is inefficient at implementing neural networks, in large part because much of it was designed for von Neumann computing schemes. Significant effort has been made towards developing electronic architectures tuned to implement artificial neural networks that exhibit improved computational speed and accuracy. Here, we propose a new architecture for a fully optical neural network that, in principle, could offer an enhancement in computational speed and power efficiency over state-of-the-art electronics for conventional inference tasks. We experimentally demonstrate the essential part of the concept using a programmable nanophotonic processor featuring a cascaded array of 56 programmable Mach–Zehnder interferometers in a silicon photonic integrated circuit and show its utility for vowel recognition. Programmable silicon nanophotonic processor empowers optical neural networks.

Subfields: Artificial Intelligence, Electrical and Electronic Engineering

View Paper: https://www.semanticscholar.org/paper/3a2746e6f559cda514fcfa5068fbd6a96069192e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Electrical and Electronic Engineering

Citations in the corpus

A light-stimulated synaptic device based on graphene hybrid phototransistor (2017, background): “By engineering our phototransistor channel to allow for evanescent light coupling with underlying waveguide, it is feasible to realize fully integrated pho tonic neuromorphic systems [45, 46].”
Large-Scale Optical Neural Networks based on Photoelectric Multiplication (2018, methodology): “This encoding, reminiscent of the Coherent Ising Machine [28, 29, 30], contrasts with other approaches used for neural networks, which encode inputs in separate spatial channels [23, 24, 22].”
Continuous-variable quantum neural networks (2018, methodology): “[31] Yichen Shen, Nicholas C Harris, Scott Skirlo, Mihika Prabhu, Tom Baehr-Jones, Michael Hochberg, Xin Sun, Shijie Zhao, Hugo Larochelle, Dirk Englund, et al.”
Space-efficient optical computing with an integrated chip diffractive neural network (2022, methodology): “Compared to the previous design with an area of 5 mm 2 and power of 3 – 7 W 7,10,13 , our IDNN chip achieves the reduced hardware footprint (0.53 mm 2 ) and low power consumption (17.5 mW), manifesting the advantage of the Fourier-based design as a scalable and power-ef ﬁ cient solution for…”
The physics of optical computing (2023, background): “…inference (as opposed to training) can load the neural-network weights into phase shifters that consume either little or no static power 1,39 and then use those weights many times by performing many inference computations with them (for example, by batching individual inferences 75 ).”
Emerging opportunities and challenges for the future of reservoir computing (2024, background): “On the other hand, development of photonic technology has brought hope for ultra-high speed and low energy consumption hardware systems, especially for neural network training 80 .”