Related references
Note: Only part of the references are listed.
Article
Multidisciplinary Sciences
Bowen Bai et al.
Summary: In this work, optical convolution is achieved through a time-wavelength plane stretching approach on a microcomb-driven chip-based photonic processing unit. Optical neural networks (ONN) have greatly benefited from parallel convolution-operation technology, which utilizes the dimension of optical wavelength to enhance complexity and functionality. However, challenges in high-level integration and on-chip operation remain for this advanced architecture. The authors develop a dedicated control and operation protocol, enabling a record high weight precision and demonstrating comparable processing capability to that of a digital computer in proof-of-concept experiments.
NATURE COMMUNICATIONS
(2023)
Article
Engineering, Electrical & Electronic
Tingyi Zhou et al.
Summary: Developing alternative machine learning approaches with low latency and small training datasets is important for applications that cannot meet the demands of deep learning methods. This article presents a new optical accelerator for AI that uses femtosecond pulses for data acquisition and computing, enabling fast classification for optical imaging, sensing, and metrology.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2022)
Article
Multidisciplinary Sciences
Logan G. Wright et al.
Summary: The study introduced a hybrid in situ-in silico algorithm called physics-aware training, which applies backpropagation to train controllable physical systems for deep physical neural networks. By demonstrating the training of diverse physical neural networks in areas like optics, mechanics, and electronics to perform audio and image classification tasks, the research showcased the universality and effectiveness of the approach.
Article
Optics
Lorenz Butschek et al.
Summary: This paper presents a photonic implementation of reservoir computing, which utilizes frequency domain multiplexing to encode neuron states. The system can process multiple neurons simultaneously, offering high-speed and high-performance advantages, and allows the implementation of output weights in the optical domain.
Article
Chemistry, Multidisciplinary
Siqi Yan et al.
Summary: This review focuses on the application of graphene in high-performance silicon photonic devices and discusses the future trends and challenges of silicon-graphene hybrid photonic devices.
APPLIED SCIENCES-BASEL
(2022)
Article
Optics
Weipeng Zhang et al.
Summary: This study experimentally demonstrates a record-high precision for photonic synapses with a dithering control scheme and numerically investigates the impact of increased synaptic precision on signal classification. It provides insights into the potential of photonic neural networks for practical applications.
Article
Multidisciplinary Sciences
Yael Gutierrez et al.
Summary: In this study, we investigated the structural change and optical contrast during the amorphous-to-crystalline transformation of antimony sulfide. We identified two crystallization regimes and proposed applications based on our findings.
Review
Engineering, Electrical & Electronic
Shengping Liu et al.
Summary: Silicon photonic platforms have significant importance in various applications and therm-optic phase shifters play a crucial role. To achieve better performance and lower costs, TOPS with various merits are needed. Different types of TOPS have been proposed and experimentally demonstrated.
FRONTIERS OF OPTOELECTRONICS
(2022)
Article
Optics
Qiyuan Yi et al.
Summary: This study demonstrates a silicon photonic flat-top 8-channel wavelength-division multiplexer (de)multiplexer for the 2 μm wavelength band. The device features low insertion loss, low channel crosstalk, and a wide operating wavelength range, making it suitable for wavelength-division multiplexing optical communication.
Article
Engineering, Electrical & Electronic
Angelina Totovic et al.
Summary: Non-von-Neumann computing architectures and deep learning training models have brought about a new era in computation, where neurons form the main architectural backbone and matrix multiplication is a fundamental tool. Neuromorphic photonics, with its energy efficient analog matrix multiply operations and advances in photonic integration, offers unique advantages in terms of speed, parallelization, and energy efficiency. However, the field has so far mainly relied on linear optical circuits and incoherent WDM approaches, without exploiting the potential of wavelengths as a new mathematical dimension. In this paper, the authors propose a new approach that combines WDM with universal linear optics, and demonstrate a high-fidelity crossbar-based neuromorphic photonic platform capable of supporting multidimensional matrix multiplication. The platform shows promising results in simulations of a convolutional neural network and offers guidelines for performance improvement.
NEUROMORPHIC COMPUTING AND ENGINEERING
(2022)
Article
Multidisciplinary Sciences
Xingyuan Xu et al.
Summary: Inspired by biological visual cortex systems, convolutional neural networks extract hierarchical features of raw data, reducing parameter complexity and improving prediction accuracy. Optical neural networks promise faster computing using broad optical bandwidths, with optical vector convolutional accelerators demonstrating efficient image processing capabilities.
Review
Engineering, Electrical & Electronic
S. Y. Siew et al.
Summary: Silicon photonics has emerged as a solution to the increasing global demand for bandwidth, leveraging the research and development in silicon semiconductor processing to enable high yield, robust processing, and low cost. Various foundries now offer commercial production services and process design kits tailored for silicon photonics integrated circuits, facilitating the development and commercialization of the technology.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2021)
Article
Optics
Alessandro Lupo et al.
Summary: The optical domain is a promising field for implementing neural networks due to its speed and parallelism. Extreme Learning Machines (ELMs) are feed-forward neural networks where only output weights are trained, while internal connections remain untrained. Experimental results show that photonic ELM performs well in classification tasks and nonlinear channel equalization tasks.
Article
Computer Science, Information Systems
Daniele Melati et al.
Article
Engineering, Electrical & Electronic
Jose David Domenech et al.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2014)
Article
Optics
Matteo Cherchi et al.
Article
Optics
Folkert Horst et al.
Article
Optics
M. C. Soriano et al.
Review
Optics
Wim Bogaerts et al.
LASER & PHOTONICS REVIEWS
(2012)
Article
Optics
Salah Ibrahim et al.
Article
Engineering, Electrical & Electronic
Joost Brouckaert et al.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2007)
Article
Physics, Applied
T Fukazawa et al.
JAPANESE JOURNAL OF APPLIED PHYSICS PART 2-LETTERS & EXPRESS LETTERS
(2004)
