An optical neural network using less than 1 photon per multiplication.
Ontology highlight
ABSTRACT:
SUBMITTER: Wang T
PROVIDER: S-EPMC8748769 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Space communication for deep-space missions, inter-satellite data transfer and Earth monitoring requires high-speed data connectivity. The reach is fundamentally dictated by the available transmission power, the aperture size, and the receiver sensitivity. A transition from radio-frequency links to optical links is now seriously being considered, as this greatly reduces the channel loss caused by diffraction. A widely studied approach uses power-efficient formats along with nanowire-based photon-counting receivers cooled to a few Kelvins operating at speeds below 1 Gb/s. However, to achieve the multi-Gb/s data rates that will be required in the future, systems relying on pre-amplified receivers together with advanced signal generation and processing techniques from fibre communications are also considered. The sensitivity of such systems is largely determined by the noise figure (NF) of the pre-amplifier, which is theoretically 3 dB for almost all amplifiers. Phase-sensitive optical amplifiers (PSAs) with their uniquely low NF of 0 dB promise to provide the best possible sensitivity for Gb/s-rate long-haul free-space links. Here, we demonstrate a novel approach using a PSA-based receiver in a free-space transmission experiment with an unprecedented bit-error-free, black-box sensitivity of 1 photon-per-information-bit (PPB) at an information rate of 10.5 Gb/s. The system adopts a simple modulation format (quadrature-phase-shift keying, QPSK), standard digital signal processing for signal recovery and forward-error correction and is straightforwardly scalable to higher data rates.
| S-EPMC7468260 | biostudies-literature
Project description:Complex-valued neural networks have many advantages over their real-valued counterparts. Conventional digital electronic computing platforms are incapable of executing truly complex-valued representations and operations. In contrast, optical computing platforms that encode information in both phase and magnitude can execute complex arithmetic by optical interference, offering significantly enhanced computational speed and energy efficiency. However, to date, most demonstrations of optical neural networks still only utilize conventional real-valued frameworks that are designed for digital computers, forfeiting many of the advantages of optical computing such as efficient complex-valued operations. In this article, we highlight an optical neural chip (ONC) that implements truly complex-valued neural networks. We benchmark the performance of our complex-valued ONC in four settings: simple Boolean tasks, species classification of an Iris dataset, classifying nonlinear datasets (Circle and Spiral), and handwriting recognition. Strong learning capabilities (i.e., high accuracy, fast convergence and the capability to construct nonlinear decision boundaries) are achieved by our complex-valued ONC compared to its real-valued counterpart.
| S-EPMC7815828 | biostudies-literature
Project description:Optical coherence tomography (OCT) is a widely used non-invasive biomedical imaging modality that can rapidly provide volumetric images of samples. Here, we present a deep learning-based image reconstruction framework that can generate swept-source OCT (SS-OCT) images using undersampled spectral data, without any spatial aliasing artifacts. This neural network-based image reconstruction does not require any hardware changes to the optical setup and can be easily integrated with existing swept-source or spectral-domain OCT systems to reduce the amount of raw spectral data to be acquired. To show the efficacy of this framework, we trained and blindly tested a deep neural network using mouse embryo samples imaged by an SS-OCT system. Using 2-fold undersampled spectral data (i.e., 640 spectral points per A-line), the trained neural network can blindly reconstruct 512 A-lines in 0.59 ms using multiple graphics-processing units (GPUs), removing spatial aliasing artifacts due to spectral undersampling, also presenting a very good match to the images of the same samples, reconstructed using the full spectral OCT data (i.e., 1280 spectral points per A-line). We also successfully demonstrate that this framework can be further extended to process 3× undersampled spectral data per A-line, with some performance degradation in the reconstructed image quality compared to 2× spectral undersampling. Furthermore, an A-line-optimized undersampling method is presented by jointly optimizing the spectral sampling locations and the corresponding image reconstruction network, which improved the overall imaging performance using less spectral data points per A-line compared to 2× or 3× spectral undersampling results. This deep learning-enabled image reconstruction approach can be broadly used in various forms of spectral-domain OCT systems, helping to increase their imaging speed without sacrificing image resolution and signal-to-noise ratio.
| S-EPMC8322159 | biostudies-literature
Project description:Optical logic operations lie at the heart of optical computing, and they enable many applications such as ultrahigh-speed information processing. However, the reported optical logic gates rely heavily on the precise control of input light signals, including their phase difference, polarization, and intensity and the size of the incident beams. Due to the complexity and difficulty in these precise controls, the two output optical logic states may suffer from an inherent instability and a low contrast ratio of intensity. Moreover, the miniaturization of optical logic gates becomes difficult if the extra bulky apparatus for these controls is considered. As such, it is desirable to get rid of these complicated controls and to achieve full logic functionality in a compact photonic system. Such a goal remains challenging. Here, we introduce a simple yet universal design strategy, capable of using plane waves as the incident signal, to perform optical logic operations via a diffractive neural network. Physically, the incident plane wave is first spatially encoded by a specific logic operation at the input layer and further decoded through the hidden layers, namely, a compound Huygens' metasurface. That is, the judiciously designed metasurface scatters the encoded light into one of two small designated areas at the output layer, which provides the information of output logic states. Importantly, after training of the diffractive neural network, all seven basic types of optical logic operations can be realized by the same metasurface. As a conceptual illustration, three logic operations (NOT, OR, and AND) are experimentally demonstrated at microwave frequencies.
| S-EPMC7154031 | biostudies-literature
Project description:Core-shells metallic nanoparticles have the advantage of possessing two plasmon resonances, one in the visible and one in the infrared part of the spectrum. This special property is used in this work to enhance the efficiency of thin film solar cells by improving the optical absorption at both wavelength ranges simultaneously by using a neural network. Although many thin-film solar cell compositions can benefit from such a design, in this work, different silver core-shell configurations were explored inside a Halide Perovskite (CH₃NH₃PbI₃) thin film. Because the number of potential configurations is infinite, only a limited number of finite difference time domain (FDTD) simulations were performed. A neural network was then trained with the simulation results to find the core-shells configurations with optimal optical absorption across different wavelength ranges. This demonstrates that core-shells nanoparticles can make an important contribution to improving solar cell performance and that neural networks can be used to find optimal results in such nanophotonic systems.
| S-EPMC6474077 | biostudies-literature
Project description:The lack of interpretability and trust is a much-criticized feature of deep neural networks. In fully connected nets, the signaling between inner layers is scrambled because backpropagation training does not require perceptrons to be arranged in any particular order. The result is a black box; this problem is particularly severe in scientific computing and digital signal processing (DSP), where neural nets perform abstract mathematical transformations that do not reduce to features or concepts. We present here a group-theoretical procedure that attempts to bring inner-layer signaling into a human-readable form, the assumption being that this form exists and has identifiable and quantifiable features-for example, smoothness or locality. We applied the proposed method to DEERNet (a DSP network used in electron spin resonance) and managed to descramble it. We found considerable internal sophistication: the network spontaneously invents a bandpass filter, a notch filter, a frequency axis rescaling transformation, frequency-division multiplexing, group embedding, spectral filtering regularization, and a map from harmonic functions into Chebyshev polynomials-in 10 min of unattended training from a random initial guess.
| S-EPMC7865153 | biostudies-literature
Project description:Data rates in optical fiber networks have increased exponentially over the past decades and core-networks are expected to operate in the peta-bit-per-second regime by 2030. As current single-mode fiber-based transmission systems are reaching their capacity limits, space-division multiplexing has been investigated as a means to increase the per-fiber capacity. Of all space-division multiplexing fibers proposed to date, multi-mode fibers have the highest spatial channel density, as signals traveling in orthogonal fiber modes share the same fiber-core. By combining a high mode-count multi-mode fiber with wideband wavelength-division multiplexing, we report a peta-bit-per-second class transmission demonstration in multi-mode fibers. This was enabled by combining three key technologies: a wideband optical comb-based transmitter to generate highly spectral efficient 64-quadrature-amplitude modulated signals between 1528 nm and 1610 nm wavelength, a broadband mode-multiplexer, based on multi-plane light conversion, and a 15-mode multi-mode fiber with optimized transmission characteristics for wideband operation.
| S-EPMC8270968 | biostudies-literature
Project description:Light scattered from multiple surfaces can be used to retrieve information of hidden environments. However, full three-dimensional retrieval of an object hidden from view by a wall has only been achieved with scanning systems and requires intensive computational processing of the retrieved data. Here we use a non-scanning, single-photon single-pixel detector in combination with a deep convolutional artificial neural network: this allows us to locate the position and to also simultaneously provide the actual identity of a hidden person, chosen from a database of people (N?=?3). Artificial neural networks applied to specific computational imaging problems can therefore enable novel imaging capabilities with hugely simplified hardware and processing times.
| S-EPMC6085360 | biostudies-literature
Project description:Interventions: None
Primary outcome(s): The primary outcome of the study is the accuracy of the CAD-CNN system for predicting histology of diminutive colorectal polyps (1-5mm) compared with the accuracy of the prediction of the endoscopist. Both the CAD-CNN system and the endoscopist will use NBI for their predictions.
Accuracy is defined as the percentage of correctly predicted optical diagnoses of the CAD-CNN system and/or endoscopist compared to the gold standard pathology. For the calculation of the accuracy, adenomas and SSLs will be dichotomised as neoplastic polyps, while HPs and other non-neoplastic histology are considered non-neoplastic.
Study Design: N/A: single arm study, Open (masking not used), N/A , unknown, Other
| 2443187 | ecrin-mdr-crc
Project description:This study examined and compared outcomes of deep learning (DL) in identifying swept-source optical coherence tomography (OCT) images without myopic macular lesions [i.e., no high myopia (nHM) vs. high myopia (HM)], and OCT images with myopic macular lesions [e.g., myopic choroidal neovascularization (mCNV) and retinoschisis (RS)]. A total of 910 SS-OCT images were included in the study as follows and analyzed by k-fold cross-validation (k = 5) using DL's renowned model, Visual Geometry Group-16: nHM, 146 images; HM, 531 images; mCNV, 122 images; and RS, 111 images (n = 910). The binary classification of OCT images with or without myopic macular lesions; the binary classification of HM images and images with myopic macular lesions (i.e., mCNV and RS images); and the ternary classification of HM, mCNV, and RS images were examined. Additionally, sensitivity, specificity, and the area under the curve (AUC) for the binary classifications as well as the correct answer rate for ternary classification were examined. The classification results of OCT images with or without myopic macular lesions were as follows: AUC, 0.970; sensitivity, 90.6%; specificity, 94.2%. The classification results of HM images and images with myopic macular lesions were as follows: AUC, 1.000; sensitivity, 100.0%; specificity, 100.0%. The correct answer rate in the ternary classification of HM images, mCNV images, and RS images were as follows: HM images, 96.5%; mCNV images, 77.9%; and RS, 67.6% with mean, 88.9%.Using noninvasive, easy-to-obtain swept-source OCT images, the DL model was able to classify OCT images without myopic macular lesions and OCT images with myopic macular lesions such as mCNV and RS with high accuracy. The study results suggest the possibility of conducting highly accurate screening of ocular diseases using artificial intelligence, which may improve the prevention of blindness and reduce workloads for ophthalmologists.
| S-EPMC7161961 | biostudies-literature