Universal linear intensity transformations using spatially incoherent diffractive processors DOI Creative Commons

Md Sadman Sakib Rahman,

Xilin Yang, Jingxi Li

et al.

Light Science & Applications, Journal Year: 2023, Volume and Issue: 12(1)

Published: Aug. 15, 2023

Abstract Under spatially coherent light, a diffractive optical network composed of structured surfaces can be designed to perform any arbitrary complex-valued linear transformation between its input and output fields-of-view (FOVs) if the total number ( N ) optimizable phase-only features is ≥~2 i o , where refer useful pixels at FOVs, respectively. Here we report design incoherent processor that approximate in time-averaged intensity FOVs. monochromatic varying point spread function H network, corresponding given, arbitrarily-selected transformation, written as m n ; ′, ′) = | h ′)| 2 same define coordinates Using numerical simulations deep learning, supervised through examples input-output profiles, demonstrate trained all-optically ≥ ~2 . We also networks for processing information multiple illumination wavelengths, operating simultaneously. Finally, numerically performs all-optical classification handwritten digits under illumination, achieving test accuracy >95%. Spatially will broadly designing visual processors work natural light.

Language: Английский

Inference in artificial intelligence with deep optics and photonics DOI
Gordon Wetzstein, Aydogan Özcan, Sylvain Gigan

et al.

Nature, Journal Year: 2020, Volume and Issue: 588(7836), P. 39 - 47

Published: Dec. 2, 2020

Language: Английский

Citations

727
Large-scale neuromorphic optoelectronic computing with a reconfigurable diffractive processing unit DOI
Tiankuang Zhou, Xing Lin, Jiamin Wu

et al.

Nature Photonics, Journal Year: 2021, Volume and Issue: 15(5), P. 367 - 373

Published: April 12, 2021

Language: Английский

Citations

480
Analogue computing with metamaterials DOI
Farzad Zangeneh‐Nejad, Dimitrios L. Sounas, Andrea Alù

et al.

Nature Reviews Materials, Journal Year: 2020, Volume and Issue: 6(3), P. 207 - 225

Published: Oct. 19, 2020

Language: Английский

Citations

343
Space-efficient optical computing with an integrated chip diffractive neural network DOI Creative Commons
Hanqing Zhu, Jun Zou, Hui Zhang

et al.

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: Feb. 24, 2022

Abstract Large-scale, highly integrated and low-power-consuming hardware is becoming progressively more important for realizing optical neural networks (ONNs) capable of advanced computing. Traditional experimental implementations need N 2 units such as Mach-Zehnder interferometers (MZIs) an input dimension to realize typical computing operations (convolutions matrix multiplication), resulting in limited scalability consuming excessive power. Here, we propose the diffractive network implementing parallel Fourier transforms, convolution application-specific using two ultracompact cells (Fourier transform operation) only MZIs. The footprint energy consumption scales linearly with data dimension, instead quadratic scaling traditional ONN framework. A ~10-fold reduction both consumption, well equal high accuracy previous MZI-based ONNs was experimentally achieved computations performed on MNIST Fashion-MNIST datasets. (IDNN) chip demonstrates a promising avenue towards scalable low-power-consumption computational chips optical-artificial-intelligence.

Language: Английский

Citations

215
Metasurface-enabled on-chip multiplexed diffractive neural networks in the visible DOI Creative Commons
Xuhao Luo, Yueqiang Hu,

Xiangnian Ou

et al.

Light Science & Applications, Journal Year: 2022, Volume and Issue: 11(1)

Published: May 27, 2022

Replacing electrons with photons is a compelling route toward high-speed, massively parallel, and low-power artificial intelligence computing. Recently, diffractive networks composed of phase surfaces were trained to perform machine learning tasks through linear optical transformations. However, the existing architectures often comprise bulky components and, most critically, they cannot mimic human brain for multitasking. Here, we demonstrate multi-skilled neural network based on metasurface device, which can on-chip multi-channel sensing multitasking in visible. The polarization multiplexing scheme subwavelength nanostructures applied construct classifier framework simultaneous recognition digital fashionable items. areal density neurons reach up 6.25 × 106 mm-2 multiplied by number channels. integrated mature complementary metal-oxide semiconductor imaging sensor, providing chip-scale architecture process information directly at physical layers energy-efficient ultra-fast image processing vision, autonomous driving, precision medicine.

Language: Английский

Citations

190
Machine learning and computation-enabled intelligent sensor design DOI Open Access
Zachary S. Ballard, Calvin Brown, Asad M. Madni

et al.

Nature Machine Intelligence, Journal Year: 2021, Volume and Issue: 3(7), P. 556 - 565

Published: June 28, 2021

Language: Английский

Citations

166
Spectrally encoded single-pixel machine vision using diffractive networks DOI Creative Commons
Jingxi Li, Deniz Mengü, Nezih Tolga Yardimci

et al.

Science Advances, Journal Year: 2021, Volume and Issue: 7(13)

Published: March 26, 2021

Diffractive networks encode the spatial information of objects into power spectrum to classify images with a single-pixel detector.

Language: Английский

Citations

165
A Review of Optical Neural Networks DOI Creative Commons
Xiubao Sui, Qiuhao Wu, Jia Liu

et al.

IEEE Access, Journal Year: 2020, Volume and Issue: 8, P. 70773 - 70783

Published: Jan. 1, 2020

Optical neural network can process information in parallel by using the technology based on free-space and integrated platform. Over last half century, development of circuits has been limited Moore's law. We know that is digital computer for successive calculation, most which cannot be made into real-time processing system. Therefore, it necessary to develop ONN device miniaturization. In this paper, we review progress optical networks. Firstly, principle artificial networks, elaborate essence matrix multiplier linear operation. Then introduce achieved interconnection waveguide interconnection. Finally talk about nonlinearity With gradual maturity nanotechnology rapid advancement silicon photonic circuits, promoted. construction future platform potential application value.

Language: Английский

Citations

155
Tackling Photonic Inverse Design with Machine Learning DOI Creative Commons
Zhaocheng Liu, Dayu Zhu, Lakshmi Raju

et al.

Advanced Science, Journal Year: 2021, Volume and Issue: 8(5)

Published: Jan. 7, 2021

Abstract Machine learning, as a study of algorithms that automate prediction and decision‐making based on complex data, has become one the most effective tools in artificial intelligence. In recent years, scientific communities have been gradually merging data‐driven approaches with research, enabling dramatic progress revealing underlying mechanisms, predicting essential properties, discovering unconventional phenomena. It is becoming an indispensable tool fields of, for instance, quantum physics, organic chemistry, medical imaging. Very recently, machine learning adopted research photonics optics alternative approach to address inverse design problem. this report, fast advances machine‐learning‐enabled photonic strategies past few years are summarized. particular, deep methods, subset algorithms, dealing intractable high degrees‐of‐freedom structure focused upon.

Language: Английский

Citations

155
Terahertz pulse shaping using diffractive surfaces DOI Creative Commons

Muhammed Veli,

Deniz Mengü, Nezih Tolga Yardimci

et al.

Nature Communications, Journal Year: 2021, Volume and Issue: 12(1)

Published: Jan. 4, 2021

Abstract Recent advances in deep learning have been providing non-intuitive solutions to various inverse problems optics. At the intersection of machine and optics, diffractive networks merge wave-optics with design task-specific elements all-optically perform tasks such as object classification vision. Here, we present a network, which is used shape an arbitrary broadband pulse into desired optical waveform, forming compact passive engineering system. We demonstrate synthesis different pulses by designing layers that collectively engineer temporal waveform input terahertz pulse. Our results direct shaping spectrum, where amplitude phase wavelengths are independently controlled through device, without need for external pump. Furthermore, physical transfer approach presented illustrate pulse-width tunability replacing part existing network newly trained layers, demonstrating its modularity. This learning-based framework can find broad applications e.g., communications, ultra-fast imaging spectroscopy.

Language: Английский

Citations

148