Showing papers by "Beihang University published in 2019"

Federated Machine Learning: Concept and Applications

Abstract: Today’s artificial intelligence still faces two major challenges. One is that, in most industries, data exists in the form of isolated islands. The other is the strengthening of data privacy and security. We propose a possible solution to these challenges: secure federated learning. Beyond the federated-learning framework first proposed by Google in 2016, we introduce a comprehensive secure federated-learning framework, which includes horizontal federated learning, vertical federated learning, and federated transfer learning. We provide definitions, architectures, and applications for the federated-learning framework, and provide a comprehensive survey of existing works on this subject. In addition, we propose building data networks among organizations based on federated mechanisms as an effective solution to allowing knowledge to be shared without compromising user privacy.

Digital Twin in Industry: State-of-the-Art

Abstract: Digital twin (DT) is one of the most promising enabling technologies for realizing smart manufacturing and Industry 4.0. DTs are characterized by the seamless integration between the cyber and physical spaces. The importance of DTs is increasingly recognized by both academia and industry. It has been almost 15 years since the concept of the DT was initially proposed. To date, many DT applications have been successfully implemented in different industries, including product design, production, prognostics and health management, and some other fields. However, at present, no paper has focused on the review of DT applications in industry. In an effort to understand the development and application of DTs in industry, this paper thoroughly reviews the state-of-the-art of the DT research concerning the key components of DTs, the current development of DTs, and the major DT applications in industry. This paper also outlines the current challenges and some possible directions for future work.

Federated Machine Learning: Concept and Applications

Abstract: Today's AI still faces two major challenges. One is that in most industries, data exists in the form of isolated islands. The other is the strengthening of data privacy and security. We propose a possible solution to these challenges: secure federated learning. Beyond the federated learning framework first proposed by Google in 2016, we introduce a comprehensive secure federated learning framework, which includes horizontal federated learning, vertical federated learning and federated transfer learning. We provide definitions, architectures and applications for the federated learning framework, and provide a comprehensive survey of existing works on this subject. In addition, we propose building data networks among organizations based on federated mechanisms as an effective solution to allow knowledge to be shared without compromising user privacy.

Rational molecular passivation for high-performance perovskite light-emitting diodes

Abstract: A major efficiency limit for solution-processed perovskite optoelectronic devices, for example light-emitting diodes, is trap-mediated non-radiative losses. Defect passivation using organic molecul ...

Fe–N–C electrocatalyst with dense active sites and efficient mass transport for high-performance proton exchange membrane fuel cells

Abstract: To achieve the US Department of Energy 2018 target set for platinum-group metal-free catalysts (PGM-free catalysts) in proton exchange membrane fuel cells, the low density of active sites must be overcome. Here, we report a class of concave Fe–N–C single-atom catalysts possessing an enhanced external surface area and mesoporosity that meets the 2018 PGM-free catalyst activity target, and a current density of 0.047 A cm–2 at 0.88 ViR-free under 1.0 bar H2–O2. This performance stems from the high density of active sites, which is realized through exposing inaccessible Fe–N4 moieties (that is, increasing their utilization) and enhancing the mass transport of the catalyst layer. Further, we establish structure–property correlations that provide a route for designing highly efficient PGM-free catalysts for practical application, achieving a power density of 1.18 W cm−2 under 2.5 bar H2–O2, and an activity of 129 mA cm−2 at 0.8 ViR-free under 1.0 bar H2–air. Iron single-atom catalysts are among the most promising fuel cell cathode materials in acid electrolyte solution. Now, Shui, Xu and co-workers report concave-shaped Fe–N–C nanoparticles with increased availability of active sites and improved mass transport, meeting the US Department of Energy 2018 target for platinum-group metal-free fuel cell catalysts.

Object Detection in 20 Years: A Survey

Abstract: Object detection, as of one the most fundamental and challenging problems in computer vision, has received great attention in recent years. Its development in the past two decades can be regarded as an epitome of computer vision history. If we think of today's object detection as a technical aesthetics under the power of deep learning, then turning back the clock 20 years we would witness the wisdom of cold weapon era. This paper extensively reviews 400+ papers of object detection in the light of its technical evolution, spanning over a quarter-century's time (from the 1990s to 2019). A number of topics have been covered in this paper, including the milestone detectors in history, detection datasets, metrics, fundamental building blocks of the detection system, speed up techniques, and the recent state of the art detection methods. This paper also reviews some important detection applications, such as pedestrian detection, face detection, text detection, etc, and makes an in-deep analysis of their challenges as well as technical improvements in recent years.

Carbon-Based Metal-Free ORR Electrocatalysts for Fuel Cells: Past, Present, and Future.

Abstract: Replacing precious platinum with earth-abundant materials for the oxygen reduction reaction (ORR) in fuel cells has been the objective worldwide for several decades. In the last 10 years, the fastest-growing branch in this area has been carbon-based metal-free ORR electrocatalysts. Great progress has been made in promoting the performance and understanding the underlying fundamentals. Here, a comprehensive review of this field is presented by emphasizing the emerging issues including the predictive design and controllable construction of porous structures and doping configurations, mechanistic understanding from the model catalysts, integrated experimental and theoretical studies, and performance evaluation in full cells. Centering on these topics, the most up-to-date results are presented, along with remarks and perspectives for the future development of carbon-based metal-free ORR electrocatalysts.

The sixth visual object tracking VOT2018 challenge results

Abstract: The Visual Object Tracking challenge VOT2018 is the sixth annual tracker benchmarking activity organized by the VOT initiative. Results of over eighty trackers are presented; many are state-of-the-art trackers published at major computer vision conferences or in journals in the recent years. The evaluation included the standard VOT and other popular methodologies for short-term tracking analysis and a “real-time” experiment simulating a situation where a tracker processes images as if provided by a continuously running sensor. A long-term tracking subchallenge has been introduced to the set of standard VOT sub-challenges. The new subchallenge focuses on long-term tracking properties, namely coping with target disappearance and reappearance. A new dataset has been compiled and a performance evaluation methodology that focuses on long-term tracking capabilities has been adopted. The VOT toolkit has been updated to support both standard short-term and the new long-term tracking subchallenges. Performance of the tested trackers typically by far exceeds standard baselines. The source code for most of the trackers is publicly available from the VOT page. The dataset, the evaluation kit and the results are publicly available at the challenge website (http://votchallenge.net).

Digital twin-driven product design framework

Abstract: With the advent of new generation information technologies in industry and product design, the big data-driven product design era has arrived. However, the big data-driven product design mainly pla...

Salient Object Detection: A Survey

Abstract: Detecting and segmenting salient objects from natural scenes, often referred to as salient object detection, has attracted great interest in computer vision. While many models have been proposed and several applications have emerged, a deep understanding of achievements and issues remains lacking. We aim to provide a comprehensive review of recent progress in salient object detection and situate this field among other closely related areas such as generic scene segmentation, object proposal generation, and saliency for fixation prediction. Covering 228 publications, we survey i) roots, key concepts, and tasks, ii) core techniques and main modeling trends, and iii) datasets and evaluation metrics for salient object detection. We also discuss open problems such as evaluation metrics and dataset bias in model performance, and suggest future research directions.

Hierarchical Nanoassembly of MoS2/Co9S8/Ni3S2/Ni as a Highly Efficient Electrocatalyst for Overall Water Splitting in a Wide pH Range.

Abstract: The design of low-cost yet high-efficiency electrocatalysts for hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) over a wide pH range is highly challenging. We now report a hierarchical co-assembly of interacting MoS2 and Co9S8 nanosheets attached on Ni3S2 nanorod arrays which are supported on nickel foam (NF). This tiered structure endows high performance toward HER and OER over a very broad pH range. By adjusting the molar ratio of the Co:Mo precursors, we have created CoMoNiS-NF- xy composites ( x: y means Co:Mo molar ratios ranging from 5:1 to 1:3) with controllable morphology and composition. The three-dimensional composites have an abundance of active sites capable of universal pH catalytic HER and OER activity. The CoMoNiS-NF-31 demonstrates the best electrocatalytic activity, giving ultralow overpotentials (113, 103, and 117 mV for HER and 166, 228, and 405 mV for OER) to achieve a current density of 10 mA cm-2 in alkaline, acidic, and neutral electrolytes, respectively. It also shows a remarkable balance between electrocatalytic activity and stability. Based on the distinguished catalytic performance of CoMoNiS-NF-31 toward HER and OER, we demonstrate a two-electrode electrolyzer performing water electrolysis over a wide pH range, with low cell voltages of 1.54, 1.45, and 1.80 V at 10 mA cm-2 in alkaline, acidic, and neutral media, respectively. First-principles calculations suggest that the high OER activity arises from electron transfer from Co9S8 to MoS2 at the interface, which alters the binding energies of adsorbed species and decreases overpotentials. Our results demonstrate that hierarchical metal sulfides can serve as highly efficient all-pH (pH = 0-14) electrocatalysts for overall water splitting.

The Role of Polarization in Photocatalysis

Abstract: Semiconductor photocatalysis as a desirable technology shows great potential in environmental remediation and renewable energy generation, but its efficiency is severely restricted by the rapid recombination of charge carriers in the bulk phase and on the surface of photocatalysts. Polarization has emerged as one of the most effective strategies for addressing the above-mentioned issues, thus effectively promoting photocatalysis. This review summarizes the recent advances on improvements of photocatalytic activity by polarization-promoted bulk and surface charge separation. Highlighted is the recent progress in charge separation advanced by different types of polarization, such as macroscopic polarization, piezoelectric polarization, ferroelectric polarization, and surface polarization, and the related mechanisms. Finally, the strategies and challenges for polarization enhancement to further enhance charge separation and photocatalysis are discussed.

The Applications of Radiomics in Precision Diagnosis and Treatment of Oncology: Opportunities and Challenges.

Abstract: Medical imaging can assess the tumor and its environment in their entirety, which makes it suitable for monitoring the temporal and spatial characteristics of the tumor. Progress in computational methods, especially in artificial intelligence for medical image process and analysis, has converted these images into quantitative and minable data associated with clinical events in oncology management. This concept was first described as radiomics in 2012. Since then, computer scientists, radiologists, and oncologists have gravitated towards this new tool and exploited advanced methodologies to mine the information behind medical images. On the basis of a great quantity of radiographic images and novel computational technologies, researchers developed and validated radiomic models that may improve the accuracy of diagnoses and therapy response assessments. Here, we review the recent methodological developments in radiomics, including data acquisition, tumor segmentation, feature extraction, and modelling, as well as the rapidly developing deep learning technology. Moreover, we outline the main applications of radiomics in diagnosis, treatment planning and evaluations in the field of oncology with the aim of developing quantitative and personalized medicine. Finally, we discuss the challenges in the field of radiomics and the scope and clinical applicability of these methods.

Combined measurements of Higgs boson couplings in proton–proton collisions at √s=13Te

Abstract: Combined measurements of the production and decay rates of the Higgs boson, as well as its couplings to vector bosons and fermions, are presented. The analysis uses the LHC proton–proton collision data set recorded with the CMS detector in 2016 at $\sqrt{s}=13\,\text {Te}\text {V} $ , corresponding to an integrated luminosity of 35.9 ${\,\text {fb}^{-1}} $ . The combination is based on analyses targeting the five main Higgs boson production mechanisms (gluon fusion, vector boson fusion, and associated production with a $\mathrm {W}$ or $\mathrm {Z}$ boson, or a top quark-antiquark pair) and the following decay modes: $\mathrm {H} \rightarrow \gamma \gamma $ , $\mathrm {Z}\mathrm {Z}$ , $\mathrm {W}\mathrm {W}$ , $\mathrm {\tau }\mathrm {\tau }$ , $\mathrm {b} \mathrm {b} $ , and $\mathrm {\mu }\mathrm {\mu }$ . Searches for invisible Higgs boson decays are also considered. The best-fit ratio of the signal yield to the standard model expectation is measured to be $\mu =1.17\pm 0.10$ , assuming a Higgs boson mass of $125.09\,\text {Ge}\text {V} $ . Additional results are given for various assumptions on the scaling behavior of the production and decay modes, including generic parametrizations based on ratios of cross sections and branching fractions or couplings. The results are compatible with the standard model predictions in all parametrizations considered. In addition, constraints are placed on various two Higgs doublet models.

Towards Optimal Structured CNN Pruning via Generative Adversarial Learning

Abstract: Structured pruning of filters or neurons has received increased focus for compressing convolutional neural networks. Most existing methods rely on multi-stage optimizations in a layer-wise manner for iteratively pruning and retraining which may not be optimal and may be computation intensive. Besides, these methods are designed for pruning a specific structure, such as filter or block structures without jointly pruning heterogeneous structures. In this paper, we propose an effective structured pruning approach that jointly prunes filters as well as other structures in an end-to-end manner. To accomplish this, we first introduce a soft mask to scale the output of these structures by defining a new objective function with sparsity regularization to align the output of baseline and network with this mask. We then effectively solve the optimization problem by generative adversarial learning (GAL), which learns a sparse soft mask in a label-free and an end-to-end manner. By forcing more scale factors in the soft mask to zero, the fast iterative shrinkage-thresholding algorithm (FISTA) can be leveraged to fast and reliably remove the corresponding structures. Extensive experiments demonstrate the effectiveness of GAL on different datasets, including MNIST, CIFAR-10 and ImageNet ILSVRC 2012. For example, on ImageNet ILSVRC 2012, the pruned ResNet-50 achieves 10.88% Top-5 error and results in a factor of 3.7x speedup. This significantly outperforms state-of-the-art methods.

Structured Knowledge Distillation for Semantic Segmentation

Abstract: In this paper, we investigate the issue of knowledge distillation for training compact semantic segmentation networks by making use of cumbersome networks. We start from the straightforward scheme, pixel-wise distillation, which applies the distillation scheme originally introduced for image classification and performs knowledge distillation for each pixel separately. We further propose to distill the structured knowledge from cumbersome networks into compact networks, which is motivated by the fact that semantic segmentation is a structured prediction problem. We study two such structured distillation schemes: (i) pair-wise distillation that distills the pairwise similarities, and (ii) holistic distillation that uses adversarial training to distill holistic knowledge. The effectiveness of our knowledge distillation approaches is demonstrated by extensive experiments on three scene parsing datasets: Cityscapes, Camvid and ADE20K.

A review of oxygen reduction mechanisms for metal-free carbon-based electrocatalysts

Abstract: The sluggish kinetics of Oxygen Reduction Reaction (ORR) at the cathode in proton exchange membrane fuel cells or metal-air batteries requires highly effective and stable electrocatalysts to boost the reaction. The low abundance and high price of Pt-based electrocatalysts hamper the widespread application of proton exchange membrane fuel cells and metal-air batteries. As promising alternatives, metal-free carbon materials, especially upon doping heteroatoms or creating defects demonstrated excellent ORR activity, which is as efficient as or even superior to commercial platinum on carbon. Significant progress on the development of advanced carbon materials as highly stable and durable catalysts has been achieved, but the catalytic mechanisms of these materials still remain undistinguished. In present review, we summarized the up-to-date progress in the studies of carbon materials, and emphasized on the combination of experiment and theory to clarify the underlying mechanisms of these materials. At last, we proposed the perspectives on the proper strategies of elucidating the mechanisms of carbon materials as electrocatalysts towards ORR.

FFA-Net: Feature Fusion Attention Network for Single Image Dehazing

Abstract: In this paper, we propose an end-to-end feature fusion at-tention network (FFA-Net) to directly restore the haze-free image. The FFA-Net architecture consists of three key components: 1) A novel Feature Attention (FA) module combines Channel Attention with Pixel Attention mechanism, considering that different channel-wise features contain totally different weighted information and haze distribution is uneven on the different image pixels. FA treats different features and pixels unequally, which provides additional flexibility in dealing with different types of information, expanding the representational ability of CNNs. 2) A basic block structure consists of Local Residual Learning and Feature Attention, Local Residual Learning allowing the less important information such as thin haze region or low-frequency to be bypassed through multiple local residual connections, let main network architecture focus on more effective information. 3) An Attention-based different levels Feature Fusion (FFA) structure, the feature weights are adaptively learned from the Feature Attention (FA) module, giving more weight to important features. This structure can also retain the information of shallow layers and pass it into deep layers. The experimental results demonstrate that our proposed FFA-Net surpasses previous state-of-the-art single image dehazing methods by a very large margin both quantitatively and qualitatively, boosting the best published PSNR metric from 30.23db to 36.39db on the SOTS indoor test dataset. Code has been made available at GitHub.

A novel neural source code representation based on abstract syntax tree

Abstract: Exploiting machine learning techniques for analyzing programs has attracted much attention. One key problem is how to represent code fragments well for follow-up analysis. Traditional information retrieval based methods often treat programs as natural language texts, which could miss important semantic information of source code. Recently, state-of-the-art studies demonstrate that abstract syntax tree (AST) based neural models can better represent source code. However, the sizes of ASTs are usually large and the existing models are prone to the long-term dependency problem. In this paper, we propose a novel AST-based Neural Network (ASTNN) for source code representation. Unlike existing models that work on entire ASTs, ASTNN splits each large AST into a sequence of small statement trees, and encodes the statement trees to vectors by capturing the lexical and syntactical knowledge of statements. Based on the sequence of statement vectors, a bidirectional RNN model is used to leverage the naturalness of statements and finally produce the vector representation of a code fragment. We have applied our neural network based source code representation method to two common program comprehension tasks: source code classification and code clone detection. Experimental results on the two tasks indicate that our model is superior to state-of-the-art approaches.

Wannier90 as a community code: new features and applications

Abstract: Wannier90 is an open-source computer program for calculating maximally-localised Wannier functions (MLWFs) from a set of Bloch states. It is interfaced to many widely used electronic-structure codes thanks to its independence from the basis sets representing these Bloch states. In the past few years the development of Wannier90 has transitioned to a community-driven model; this has resulted in a number of new developments that have been recently released in Wannier90 v3.0. In this article we describe these new functionalities, that include the implementation of new features for wannierisation and disentanglement (symmetry-adapted Wannier functions, selectively-localised Wannier functions, selected columns of the density matrix) and the ability to calculate new properties (shift currents and Berry-curvature dipole, and a new interface to many-body perturbation theory); performance improvements, including parallelisation of the core code; enhancements in functionality (support for spinor-valued Wannier functions, more accurate methods to interpolate quantities in the Brillouin zone); improved usability (improved plotting routines, integration with high-throughput automation frameworks), as well as the implementation of modern software engineering practices (unit testing, continuous integration, and automatic source-code documentation). These new features, capabilities, and code development model aim to further sustain and expand the community uptake and range of applicability, that nowadays spans complex and accurate dielectric, electronic, magnetic, optical, topological and transport properties of materials.

A bionic stretchable nanogenerator for underwater sensing and energy harvesting.

Abstract: Soft wearable electronics for underwater applications are of interest, but depend on the development of a waterproof, long-term sustainable power source. In this work, we report a bionic stretchable nanogenerator for underwater energy harvesting that mimics the structure of ion channels on the cytomembrane of electrocyte in an electric eel. Combining the effects of triboelectrification caused by flowing liquid and principles of electrostatic induction, the bionic stretchable nanogenerator can harvest mechanical energy from human motion underwater and output an open-circuit voltage over 10 V. Underwater applications of a bionic stretchable nanogenerator have also been demonstrated, such as human body multi-position motion monitoring and an undersea rescue system. The advantages of excellent flexibility, stretchability, outstanding tensile fatigue resistance (over 50,000 times) and underwater performance make the bionic stretchable nanogenerator a promising sustainable power source for the soft wearable electronics used underwater.

Symbiotic cardiac pacemaker.

Abstract: Self-powered implantable medical electronic devices that harvest biomechanical energy from cardiac motion, respiratory movement and blood flow are part of a paradigm shift that is on the horizon. Here, we demonstrate a fully implanted symbiotic pacemaker based on an implantable triboelectric nanogenerator, which achieves energy harvesting and storage as well as cardiac pacing on a large-animal scale. The symbiotic pacemaker successfully corrects sinus arrhythmia and prevents deterioration. The open circuit voltage of an implantable triboelectric nanogenerator reaches up to 65.2 V. The energy harvested from each cardiac motion cycle is 0.495 μJ, which is higher than the required endocardial pacing threshold energy (0.377 μJ). Implantable triboelectric nanogenerators for implantable medical devices offer advantages of excellent output performance, high power density, and good durability, and are expected to find application in fields of treatment and diagnosis as in vivo symbiotic bioelectronics. Implantable medical electronic devices are limited by battery lifetime and inflexibility, but self-powered devices can harvest biomechanical energy. Here the authors demonstrate cardiac pacing and correction of sinus arrhythmia with a symbiotic cardiac pacemaker, which is an implanted self-powered pacing system powered by cardiac motion, in a swine.

Differentiable Soft Quantization: Bridging Full-Precision and Low-Bit Neural Networks

Abstract: Hardware-friendly network quantization (e.g., binary/uniform quantization) can efficiently accelerate the inference and meanwhile reduce memory consumption of the deep neural networks, which is crucial for model deployment on resource-limited devices like mobile phones. However, due to the discreteness of low-bit quantization, existing quantization methods often face the unstable training process and severe performance degradation. To address this problem, in this paper we propose Differentiable Soft Quantization (DSQ) to bridge the gap between the full-precision and low-bit networks. DSQ can automatically evolve during training to gradually approximate the standard quantization. Owing to its differentiable property, DSQ can help pursue the accurate gradients in backward propagation, and reduce the quantization loss in forward process with an appropriate clipping range. Extensive experiments over several popular network structures show that training low-bit neural networks with DSQ can consistently outperform state-of-the-art quantization methods. Besides, our first efficient implementation for deploying 2 to 4-bit DSQ on devices with ARM architecture achieves up to 1.7× speed up, compared with the open-source 8-bit high-performance inference framework NCNN [31].

High thermoelectric performance in low-cost SnS 0.91 Se 0.09 crystals.

Abstract: Thermoelectric technology allows conversion between heat and electricity. Many good thermoelectric materials contain rare or toxic elements, so developing low-cost and high-performance thermoelectric materials is warranted. Here, we report the temperature-dependent interplay of three separate electronic bands in hole-doped tin sulfide (SnS) crystals. This behavior leads to synergistic optimization between effective mass (m*) and carrier mobility (μ) and can be boosted through introducing selenium (Se). This enhanced the power factor from ~30 to ~53 microwatts per centimeter per square kelvin (μW cm−1 K−2 at 300 K), while lowering the thermal conductivity after Se alloying. As a result, we obtained a maximum figure of merit ZT (ZTmax) of ~1.6 at 873 K and an average ZT (ZTave) of ~1.25 at 300 to 873 K in SnS0.91Se0.09 crystals. Our strategy for band manipulation offers a different route for optimizing thermoelectric performance. The high-performance SnS crystals represent an important step toward low-cost, Earth-abundant, and environmentally friendly thermoelectrics.

An Overall Distribution Particle Swarm Optimization MPPT Algorithm for Photovoltaic System Under Partial Shading

Abstract: Solar photovoltaic (PV) systems under partial shading conditions (PSCs) have a nonmonotonic P – V characteristic with multiple local maximum power points, which makes the existing maximum power point tracking (MPPT) algorithms unsatisfactory performance for global MPPT, if not invalid. This paper proposes a novel overall distribution (OD) MPPT algorithm to rapidly search the area near the global maximum power points, which is further integrated with the particle swarm optimization (PSO) MPPT algorithm to improve the accuracy of MPPT. Through simulations and experimentations, the higher effectiveness and accuracy of the proposed OD-PSO MPPT algorithm in solar PV systems is demonstrated in comparison to two existing artificial intelligence MPPT algorithms.