Photonic neuromorphic architecture for tens-of-task lifelong learning.

Scalable, high-capacity, and low-power computing architecture is the primary assurance for increasingly manifold and large-scale machine learning tasks. Traditional electronic artificial agents by conventional power-hungry processors have faced the issues of energy and scaling walls, hindering them from the sustainable performance improvement and iterative multi-task learning. Referring to another modality of light, photonic computing has been progressively applied in high-efficient neuromorphic systems. Here, we innovate a reconfigurable lifelong-learning optical neural network (L2ONN), for highly-integrated tens-of-task machine intelligence with elaborated algorithm-hardware co-design. Benefiting from the inherent sparsity and parallelism in massive photonic connections, L2ONN learns each single task by adaptively activating sparse photonic neuron connections in the coherent light field, while incrementally acquiring expertise on various tasks by gradually enlarging the activation. The multi-task optical features are parallelly processed by multi-spectrum representations allocated with different wavelengths. Extensive evaluations on free-space and on-chip architectures confirm that for the first time, L2ONN avoided the catastrophic forgetting issue of photonic computing, owning versatile skills on challenging tens-of-tasks (vision classification, voice recognition, medical diagnosis, etc.) with a single model. Particularly, L2ONN achieves more than an order of magnitude higher efficiency than the representative electronic artificial neural networks, and 14× larger capacity than existing optical neural networks while maintaining competitive performance on each individual task. The proposed photonic neuromorphic architecture points out a new form of lifelong learning scheme, permitting terminal/edge AI systems with light-speed efficiency and unprecedented scalability.

Research Progress on Replacement and Production Technology for Long-Distance Liquefied Ethane Pipelines.

Ethane is used as raw material to produce ethylene, which is the most important basic raw material for the petrochemical industry. The liquid phase ethane transportation method has the advantages of large transportation capacity and high economy. In this paper, the research progress of long-distance ethane pipelines is reviewed from the aspects of construction, phase change characteristics, standards and specifications, replacement, and production technology. The phase change characteristics of ethane-nitrogen mixed gas in the replacement process are discussed. In addition, the applicability of existing standards, specifications, and related replacement production technologies to liquefied ethane pipelines was analyzed, and operational recommendations have been given. Suggestions for future research are put forward to promote the application of pipeline replacement and production technology for ethane long transportation.

Training large-scale optoelectronic neural networks with dual-neuron optical-artificial learning.

Optoelectronic neural networks (ONN) are a promising avenue in AI computing due to their potential for parallelization, power efficiency, and speed. Diffractive neural networks, which process information by propagating encoded light through trained optical elements, have garnered interest. However, training large-scale diffractive networks faces challenges due to the computational and memory costs of optical diffraction modeling. Here, we present DANTE, a dual-neuron optical-artificial learning architecture. Optical neurons model the optical diffraction, while artificial neurons approximate the intensive optical-diffraction computations with lightweight functions. DANTE also improves convergence by employing iterative global artificial-learning steps and local optical-learning steps. In simulation experiments, DANTE successfully trains large-scale ONNs with 150 million neurons on ImageNet, previously unattainable, and accelerates training speeds significantly on the CIFAR-10 benchmark compared to single-neuron learning. In physical experiments, we develop a two-layer ONN system based on DANTE, which can effectively extract features to improve the classification of natural images.

A multichannel optical computing architecture for advanced machine vision.

Endowed with the superior computing speed and energy efficiency, optical neural networks (ONNs) have attracted ever-growing attention in recent years. Existing optical computing architectures are mainly single-channel due to the lack of advanced optical connection and interaction operators, solving simple tasks such as hand-written digit classification, saliency detection, etc. The limited computing capacity and scalability of single-channel ONNs restrict the optical implementation of advanced machine vision. Herein, we develop Monet: a multichannel optical neural network architecture for a universal multiple-input multiple-channel optical computing based on a novel projection-interference-prediction framework where the inter- and intra- channel connections are mapped to optical interference and diffraction. In our Monet, optical interference patterns are generated by projecting and interfering the multichannel inputs in a shared domain. These patterns encoding the correspondences together with feature embeddings are iteratively produced through the projection-interference process to predict the final output optically. For the first time, Monet validates that multichannel processing properties can be optically implemented with high-efficiency, enabling real-world intelligent multichannel-processing tasks solved via optical computing, including 3D/motion detections. Extensive experiments on different scenarios demonstrate the effectiveness of Monet in handling advanced machine vision tasks with comparative accuracy as the electronic counterparts yet achieving a ten-fold improvement in computing efficiency. For intelligent computing, the trends of dealing with real-world advanced tasks are irreversible. Breaking the capacity and scalability limitations of single-channel ONN and further exploring the multichannel processing potential of wave optics, we anticipate that the proposed technique will accelerate the development of more powerful optical AI as critical support for modern advanced machine vision.

GigaMVS: A Benchmark for Ultra-Large-Scale Gigapixel-Level 3D Reconstruction.

Multiview stereopsis (MVS) methods, which can reconstruct both the 3D geometry and texture from multiple images, have been rapidly developed and extensively investigated from the feature engineering methods to the data-driven ones. However, there is no dataset containing both the 3D geometry of large-scale scenes and high-resolution observations of small details to benchmark the algorithms. To this end, we present GigaMVS, the first gigapixel-image-based 3D reconstruction benchmark for ultra-large-scale scenes. The gigapixel images, with both wide field-of-view and high-resolution details, can clearly observe both the Palace-scale scene structure and Relievo-scale local details. The ground-truth geometry is captured by the laser scanner, which covers ultra-large-scale scenes with an average area of 8667 m 2 and a maximum area of 32007 m 2. Owing to the extremely large scale, complex occlusion, and gigapixel-level images, GigaMVS exposes problems that emerge from the poor scalability and efficiency of the existing MVS algorithms. We thoroughly investigate the state-of-the-art methods in terms of geometric and textural measurements, which point to the weakness of the existing methods and promising opportunities for future works. We believe that GigaMVS can benefit the community of 3D reconstruction and support the development of novel algorithms balancing robustness, scalability and accuracy.

Massively parallel functional photoacoustic computed tomography of the human brain.

Blood-oxygen-level-dependent (BOLD) functional magnetic resonance imaging of the human brain requires bulky equipment for the generation of magnetic fields. Photoacoustic computed tomography obviates the need for magnetic fields by using light and sound to measure deoxyhaemoglobin and oxyhaemoglobin concentrations to then quantify oxygen saturation and blood volumes. Yet, the available imaging speeds, fields of view (FOV), sensitivities and penetration depths have been insufficient for functional imaging of the human brain. Here, we show that massively parallel ultrasonic transducers arranged hemispherically around the human head can produce tomographic images of the brain with a 10-cm-diameter FOV and spatial and temporal resolutions of 350 µm and 2 s, respectively. In patients who had a hemicraniectomy, a comparison of functional photoacoustic computed tomography and 7 T BOLD functional magnetic resonance imaging showed a strong spatial correspondence in the same FOV and a high temporal correlation between BOLD signals and photoacoustic signals, with the latter enabling faster detection of functional activation. Our findings establish the use of photoacoustic computed tomography for human brain imaging.

The effect of surface-capping oleic acid on the optical properties of lanthanide-doped nanocrystals.

The rapid development of nanotechnology has placed a higher demand on the synthesis of nanomaterials. Benefiting from its capability to keep nanoparticles away from aggregation, oleic acid (OA) has been routinely utilized as a capping agent in the synthesis of monodisperse nanocrystals. To satisfy downstream biological applications, hydrophobic OA capping on the surface should be removed or coated, but scarce attention has been paid to its influence on the optical properties of nanocrystals. In this work, the effect of surface-capping OA has been systematically explored on the optical properties of lanthanide-doped upconversion and downshifting nanocrystals, respectively. The emission intensity and lifetime of emissive lanthanides have been compared between OA-capped and ligand-free nanocrystals either in solid state or in colloidal solution. In solid state, surface-capping OA can significantly influence both emission intensity and radiative transition possibility of emissive lanthanides. However, in colloidal solution, a distinct variation between OA-capped and ligand-free nanocrystals is observed. Besides, the effect of OA on the luminescence dynamics of lanthanides with different energy gaps (emitting level to the next-lower-energy level) has been investigated in colloidal solution. The possible mechanism for the effect of OA on the optical properties of lanthanide-doped nanocrystals has been further proposed.

High-speed three-dimensional photoacoustic computed tomography for preclinical research and clinical translation.

Photoacoustic computed tomography (PACT) has generated increasing interest for uses in preclinical research and clinical translation. However, the imaging depth, speed, and quality of existing PACT systems have previously limited the potential applications of this technology. To overcome these issues, we developed a three-dimensional photoacoustic computed tomography (3D-PACT) system that features large imaging depth, scalable field of view with isotropic spatial resolution, high imaging speed, and superior image quality. 3D-PACT allows for multipurpose imaging to reveal detailed angiographic information in biological tissues ranging from the rodent brain to the human breast. In the rat brain, we visualize whole brain vasculatures and hemodynamics. In the human breast, an in vivo imaging depth of 4 cm is achieved by scanning the breast within a single breath hold of 10 s. Here, we introduce the 3D-PACT system to provide a unique tool for preclinical research and an appealing prototype for clinical translation.

A modular hierarchical array camera.

Array cameras removed the optical limitations of a single camera and paved the way for high-performance imaging via the combination of micro-cameras and computation to fuse multiple aperture images. However, existing solutions use dense arrays of cameras that require laborious calibration and lack flexibility and practicality. Inspired by the cognition function principle of the human brain, we develop an unstructured array camera system that adopts a hierarchical modular design with multiscale hybrid cameras composing different modules. Intelligent computations are designed to collaboratively operate along both intra- and intermodule pathways. This system can adaptively allocate imagery resources to dramatically reduce the hardware cost and possesses unprecedented flexibility, robustness, and versatility. Large scenes of real-world data were acquired to perform human-centric studies for the assessment of human behaviours at the individual level and crowd behaviours at the population level requiring high-resolution long-term monitoring of dynamic wide-area scenes.

CrossNet++: Cross-Scale Large-Parallax Warping for Reference-Based Super-Resolution.

The ability of camera arrays to efficiently capture higher space-bandwidth product than single cameras has led to various multiscale and hybrid systems. These systems play vital roles in computational photography, including light field imaging, 360 VR camera, gigapixel videography, etc. One of the critical tasks in multiscale hybrid imaging is matching and fusing cross-resolution images from different cameras under perspective parallax. In this paper, we investigate the reference-based super-resolution (RefSR) problem associated with dual-camera or multi-camera systems. RefSR consists of super-resolving a low-resolution (LR) image given an external high-resolution (HR) reference image, where they suffer both a significant resolution gap ( 8×) and large parallax (  âˆ¼ 10% pixel displacement). We present CrossNet++, an end-to-end network containing novel two-stage cross-scale warping modules, image encoder and fusion decoder. The stage I learns to narrow down the parallax distinctively with the strong guidance of landmarks and intensity distribution consensus. Then the stage II operates more fine-grained alignment and aggregation in feature domain to synthesize the final super-resolved image. To further address the large parallax, new hybrid loss functions comprising warping loss, landmark loss and super-resolution loss are proposed to regularize training and enable better convergence. CrossNet++ significantly outperforms the state-of-art on light field datasets as well as real dual-camera data. We further demonstrate the generalization of our framework by transferring it to video super-resolution and video denoising.

Transcranial photoacoustic computed tomography based on a layered back-projection method.

A major challenge of transcranial human brain photoacoustic computed tomography (PACT) is correcting for the acoustic aberration induced by the skull. Here, we present a modified universal back-projection (UBP) method, termed layered UBP (L-UBP), that can de-aberrate the transcranial PA signals by accommodating the skull heterogeneity into conventional UBP. In L-UBP, the acoustic medium is divided into multiple layers: the acoustic coupling fluid layer between the skull and detectors, the skull layer, and the brain tissue layer, which are assigned different acoustic properties. The transmission coefficients and wave conversion are considered at the fluid-skull and skull-tissue interfaces. Simulations of transcranial PACT using L-UBP were conducted to validate the method. Ex vivo experiments with a newly developed three-dimensional PACT system with 1-MHz center frequency demonstrated that L-UBP can substantially improve the image quality compared to conventional UBP.

Effect of Coiling Temperature on Microstructure, Properties and Resistance to Fish-Scaling of Hot Rolled Enamel Steel.

The microstructure, mechanical properties, and hydrogen permeation behavior of hot rolled enamel steel were investigated. Three coiling temperatures were adopted to gain different sizes of ferrite grain and TiC precipitates. The results show that a large number of interphase precipitates of nano-sized TiC can be obtained at coiling temperatures of 650 and 700 °C, while a few precipitates are found in experimental steel when coiling temperature is 600 °C. The yield strength and ultimate tensile strength decrease with increasing coiling temperature, while elongation increases. The experimental steel has the best resistance to fish-scaling at coiling temperature of 700 °C, due to the large quantities of nano-sized interphase precipitates of TiC.

A novel method for aqueous synthesis of CdTe duantum dots.

We have developed a simple and an economical one-pot method to synthesize water-soluble CdTe quantum dots (QDs) using hydroxylamine hydrochloride (HAH) as reduction and l-cysteine (CYS) as the ligand. The size of the CdTe QDs could easily be controlled by the duration of reflux and monitored by absorption and photoluminescence spectra. The factors influencing the photoluminescence quantum yields (PL QYs) on the QYs of CdTe NCs were investigated and the optimum conditions were determined. Under the optimum conditions (pH=11.0, the concentration of Cd(2+) was 1.0mmolL(-1) and the molar ratio of Cd(2+):Te(2)(-):CYS:HAH was 1:0.05:2.4:5), photoluminescence quantum yields of the CdTe QDs have been improved significantly and the maximum QYs of the QDs can achieve to 47%. The QDs were characterized by Fourier transform infrared spectrometry (FTIR), transmission-electron microscopy (TEM) and X-ray powder diffraction (XRD). The XRD patterns indicated that CdS was formed in the preparation process of CdTe QDs. This CdS shell could effectively passivate the surface trap states, and enhance the PL QY and stability of the CdTe QDs.