Photonic Ising machines for combinatorial optimization problems DOI
Yuan Gao, Guanyu Chen, Qi Luo

и другие.

Applied Physics Reviews, Год журнала: 2024, Номер 11(4)

Опубликована: Окт. 14, 2024

The demand for efficient solvers of complicated combinatorial optimization problems, especially those classified as NP-complete or NP-hard, has recently led to increased exploration novel computing architectures. One prominent collective state paradigm embodied in the so-called Ising machines attracted considerable research attention due its ability optimize complex problems with large numbers interacting variables. model-inspired solvers, thus named mathematical similarities well-known model from solid-state physics, represent a promising alternative traditional von Neumann computer architectures their high degree inherent parallelism. While there are many possible physical realizations just implementations any binary computer, photonic (PIMs) use primarily optical components computation, taking advantage features like lower power consumption, fast calculation speeds, leveraging optics perform calculations themselves, possessing decent scalability and noise tolerance. Photonic form PIMs may offer certain computational advantages that not easily achieved non-photonic approaches is nonetheless an altogether fascinating application photonics computing. In this review, we provide overview generally, introducing why they useful, what types can tackle, how different be compared benchmarked. We delineate various operational mechanisms, advantages, limitations vis-à-vis machines. describe scalability, interconnectivity, performance, dimensions. As continues progress, potential could well emerge way handle challenging across diverse domains. This review serves comprehensive resource researchers practitioners interested understanding capabilities addressing such problems.

Язык: Английский

Smoothed analysis-based noise manipulation for spatial photonic Ising machines DOI
Xin Ye, Wenjia Zhang, Zuyuan He

и другие.

Chinese Optics Letters, Год журнала: 2025, Номер 23(3), С. 032501 - 032501

Опубликована: Янв. 1, 2025

Язык: Английский

Процитировано

0
High-Efficiency Bident Edge Coupler based on Thin Film Lithium Niobate DOI Creative Commons
Min Liu, Tianheng Zhang,

Binhang Xu

и другие.

Optics Express, Год журнала: 2025, Номер 33(7), С. 16338 - 16338

Опубликована: Март 19, 2025

Thin-film lithium niobate (TFLN) has gained significant attention as a platform for photonic integrated circuits (PICs) due to its exceptional properties, including high nonlinear coefficients and strong electro-optic response. However, efficient coupling between TFLN chips optical fibers remains challenge, with current edge coupler designs often facing issues related fabrication complexity alignment precision. In this paper, we propose high-efficiency, fabrication-friendly, alignment-tolerant bident based on overcome these limitations. The design comprises three distinct regions: the facet region, beam combining adiabatic taper region. It achieves measured efficiency of better than -1.52 dB per at 1550 nm, an average ranging from -1.85 -1.6 across C-band, mode field diameter approximately 4 μm. Notably, offers large tolerance ±1 This simplifies process, requiring only single overlay step. proposed practical solution interface packaging in devices, enabling performance low process complexity.

Язык: Английский

Процитировано

0
Minimalist Photonic Processor for High‐Volume and Versatile Linear Computation DOI
Zhenhua Li,

Zhaoang Deng,

Jie Liu

и другие.

Laser & Photonics Review, Год журнала: 2025, Номер unknown

Опубликована: Апрель 3, 2025

Abstract By fully exploiting the rich parameter dimensions of light wave including time, wavelength, transverse space, or mode, photonic integrated circuits potentially offer low‐latency, high‐throughput, and energy‐efficient solutions for acceleration multimodal linear data processing in artificial intelligence‐related computational tasks. However, many existing schemes tailor specific construct architectures to suit operations and, therefore not making full use optical resources lacking versatility adapting different operations. Their scale is often linked that operands, lack flexibility when dealing with variable sizes. A novel multi‐dimensional minimalist processor (MD‐MPP) architecture demonstrated, capable simultaneously scalably utilizing space multiplexing achieve high throughput, versatile operations, flexible adaption, performing all‐optical multiply‐and‐accumulate (MAC) vector dot‐products, matrix‐vector‐multiplication, single‐/multi‐kernel convolution time‐recursive, wavelength‐parallel spatial‐parallel fashions. As a verification, chip fabricated thin‐film lithium niobate (TFLN) experimentally implements multi‐wavelength optoelectronic convolutional neural networks up 36.7 billion MAC per second (or 73.4 GOPS) device underscoring its potential be promising candidate computing at volumes lower energy consumption.

Язык: Английский

Процитировано

0
Monolithically integrated asynchronous optical recurrent accelerator DOI Creative Commons
Bo Wu,

Haojun Zhou,

Junwei Cheng

и другие.

eLight, Год журнала: 2025, Номер 5(1)

Опубликована: Май 5, 2025

Язык: Английский

Процитировано

0
A monolithically integrated optical Ising machine DOI Creative Commons
Bo Wu, Wenkai Zhang,

Shiji Zhang

и другие.

Nature Communications, Год журнала: 2025, Номер 16(1)

Опубликована: Май 8, 2025

The growing demand for enhanced computational power and energy efficiency has driven the development of optical Ising machines solving combinatorial optimization problems. However, existing implementations face challenges in integration density efficiency. Here, we propose a monolithically integrated four-spin machine based on optoelectronic coupled oscillators. This system integrates custom-designed Mach-Zehnder interferometer (MZI) symmetric matrix with high-efficiency optical-electrical (OEC) nonlinear unit. OEC unit an ultra-compact 0.01 mm² footprint achieves 4 mW per unit, ensuring scalability. reconfigurable real-valued coupling mean fidelity 0.986. spin evolution time is measured as 150 ns, 1.71 ns round-trip confirmed through bandwidth measurements. successfully finds ground states various problems, demonstrating its effectiveness. work represents significant step toward monolithic all-optical physical annealing systems, minimizing footprint, consumption, convergence time.

Язык: Английский

Процитировано

0
Optoelectronic Ising Machine-Based Maximum Likelihood MIMO Detection for Enhanced SDM Fiber-Optic Transmission DOI
Yibin Wan, Zhenhua Li, Zihao Chen

и другие.

Optical Fiber Communication Conference (OFC) 2022, Год журнала: 2025, Номер unknown, С. W1D.4 - W1D.4

Опубликована: Янв. 1, 2025

An optoelectronic Ising machine-based maximum likelihood MIMO detection is proposed and experimentally demonstrated in SDM fiber-optic communication systems, achieving a 16.6% net rate improvement than conventional DSP with tolerant iteration time energy consumption.

Язык: Английский

Процитировано

0
Photonic Ising machines for combinatorial optimization problems DOI
Yuan Gao, Guanyu Chen, Qi Luo

и другие.

Applied Physics Reviews, Год журнала: 2024, Номер 11(4)

Опубликована: Окт. 14, 2024

The demand for efficient solvers of complicated combinatorial optimization problems, especially those classified as NP-complete or NP-hard, has recently led to increased exploration novel computing architectures. One prominent collective state paradigm embodied in the so-called Ising machines attracted considerable research attention due its ability optimize complex problems with large numbers interacting variables. model-inspired solvers, thus named mathematical similarities well-known model from solid-state physics, represent a promising alternative traditional von Neumann computer architectures their high degree inherent parallelism. While there are many possible physical realizations just implementations any binary computer, photonic (PIMs) use primarily optical components computation, taking advantage features like lower power consumption, fast calculation speeds, leveraging optics perform calculations themselves, possessing decent scalability and noise tolerance. Photonic form PIMs may offer certain computational advantages that not easily achieved non-photonic approaches is nonetheless an altogether fascinating application photonics computing. In this review, we provide overview generally, introducing why they useful, what types can tackle, how different be compared benchmarked. We delineate various operational mechanisms, advantages, limitations vis-à-vis machines. describe scalability, interconnectivity, performance, dimensions. As continues progress, potential could well emerge way handle challenging across diverse domains. This review serves comprehensive resource researchers practitioners interested understanding capabilities addressing such problems.

Язык: Английский

Процитировано

2