Showing papers by "Xi'an Jiaotong University published in 2015"

Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification

Abstract: Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on our PReLU networks (PReLU-nets), we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66%). To our knowledge, our result is the first to surpass human-level performance (5.1%, Russakovsky et al.) on this visual recognition challenge.

Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification

Abstract: Rectified activation units (rectifiers) are essential for state-of-the-art neural networks. In this work, we study rectifier neural networks for image classification from two aspects. First, we propose a Parametric Rectified Linear Unit (PReLU) that generalizes the traditional rectified unit. PReLU improves model fitting with nearly zero extra computational cost and little overfitting risk. Second, we derive a robust initialization method that particularly considers the rectifier nonlinearities. This method enables us to train extremely deep rectified models directly from scratch and to investigate deeper or wider network architectures. Based on the learnable activation and advanced initialization, we achieve 4.94% top-5 test error on the ImageNet 2012 classification dataset. This is a 26% relative improvement over the ILSVRC 2014 winner (GoogLeNet, 6.66% [33]). To our knowledge, our result is the first to surpass the reported human-level performance (5.1%, [26]) on this dataset.

Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition

Abstract: Existing deep convolutional neural networks (CNNs) require a fixed-size (e.g., 224 $\times$ 224) input image. This requirement is “artificial” and may reduce the recognition accuracy for the images or sub-images of an arbitrary size/scale. In this work, we equip the networks with another pooling strategy, “spatial pyramid pooling”, to eliminate the above requirement. The new network structure, called SPP-net, can generate a fixed-length representation regardless of image size/scale. Pyramid pooling is also robust to object deformations. With these advantages, SPP-net should in general improve all CNN-based image classification methods. On the ImageNet 2012 dataset, we demonstrate that SPP-net boosts the accuracy of a variety of CNN architectures despite their different designs. On the Pascal VOC 2007 and Caltech101 datasets, SPP-net achieves state-of-the-art classification results using a single full-image representation and no fine-tuning. The power of SPP-net is also significant in object detection. Using SPP-net, we compute the feature maps from the entire image only once, and then pool features in arbitrary regions (sub-images) to generate fixed-length representations for training the detectors. This method avoids repeatedly computing the convolutional features. In processing test images, our method is 24-102 $\times$ faster than the R-CNN method, while achieving better or comparable accuracy on Pascal VOC 2007. In ImageNet Large Scale Visual Recognition Challenge (ILSVRC) 2014, our methods rank #2 in object detection and #3 in image classification among all 38 teams. This manuscript also introduces the improvement made for this competition.

Isothermal Amplification of Nucleic Acids

Abstract: Isothermal amplification of nucleic acids is a simple process that rapidly and efficiently accumulates nucleic acid sequences at constant temperature. Since the early 1990s, various isothermal amplification techniques have been developed as alternatives to polymerase chain reaction (PCR). These isothermal amplification methods have been used for biosensing targets such as DNA, RNA, cells, proteins, small molecules, and ions. The applications of these techniques for in situ or intracellular bioimaging and sequencing have been amply demonstrated. Amplicons produced by isothermal amplification methods have also been utilized to construct versatile nucleic acid nanomaterials for promising applications in biomedicine, bioimaging, and biosensing. The integration of isothermal amplification into microsystems or portable devices improves nucleic acid-based on-site assays and confers high sensitivity. Single-cell and single-molecule analyses have also been implemented based on integrated microfluidic systems. In this review, we provide a comprehensive overview of the isothermal amplification of nucleic acids encompassing work published in the past two decades. First, different isothermal amplification techniques are classified into three types based on reaction kinetics. Then, we summarize the applications of isothermal amplification in bioanalysis, diagnostics, nanotechnology, materials science, and device integration. Finally, several challenges and perspectives in the field are discussed.

A Large‐Bandgap Conjugated Polymer for Versatile Photovoltaic Applications with High Performance

Abstract: A new copolymer PM6 based on fluorothienyl-substituted benzodithiophene is synthesized and characterized. The inverted polymer solar cells based on PM6 exhibit excellent performance with Voc of 0.98 V and power conversion efficiency (PCE) of 9.2% for a thin-film thickness of 75 nm. Furthermore, the single-junction semitransparent device shows a high PCE of 5.7%.

Sparse whole-genome sequencing identifies two loci for major depressive disorder

Abstract: Genomic analysis of 5,303 Chinese women with recurrent major depressive disorder (MDD) enables the identification and replication of two genome-wide significant loci contributing to risk of MDD on chromosome 10: one near the SIRT1 gene; the other in an intron of the LHPP gene.

Functionalization of phosphorescent emitters and their host materials by main-group elements for phosphorescent organic light-emitting devices.

Abstract: Phosphorescent organic light-emitting devices (OLEDs) have attracted increased attention from both academic and industrial communities due to their potential practical application in high-resolution full-color displays and energy-saving solid-state lightings. The performance of phosphorescent OLEDs is mainly limited by the phosphorescent transition metal complexes (such as iridium(III), platinum(II), gold(III), ruthenium(II), copper(I) and osmium(II) complexes, etc.) which can play a crucial role in furnishing efficient energy transfer, balanced charge injection/transporting character and high quantum efficiency in the devices. It has been shown that functionalized main-group element (such as boron, silicon, nitrogen, phosphorus, oxygen, sulfur and fluorine, etc.) moieties can be incorporated into phosphorescent emitters and their host materials to tune their triplet energies, frontier molecular orbital energies, charge injection/transporting behavior, photophysical properties and thermal stability and hence bring about highly efficient phosphorescent OLEDs. So, in this review, the recent advances in the phosphorescent emitters and their host materials functionalized with various main-group moieties will be introduced from the point of view of their structure-property relationship. The main emphasis lies on the important role played by the main-group element groups in addressing the key issues of both phosphorescent emitters and their host materials to fulfill high-performance phosphorescent OLEDs.

Learning a convolutional neural network for non-uniform motion blur removal

Abstract: In this paper, we address the problem of estimating and removing non-uniform motion blur from a single blurry image. We propose a deep learning approach to predicting the probabilistic distribution of motion blur at the patch level using a convolutional neural network (CNN). We further extend the candidate set of motion kernels predicted by the CNN using carefully designed image rotations. A Markov random field model is then used to infer a dense non-uniform motion blur field enforcing motion smoothness. Finally, motion blur is removed by a non-uniform deblurring model using patch-level image prior. Experimental evaluations show that our approach can effectively estimate and remove complex non-uniform motion blur that is not handled well by previous approaches.

Neutrino Physics with JUNO

Abstract: The Jiangmen Underground Neutrino Observatory (JUNO), a 20 kton multi-purpose underground liquid scintillator detector, was proposed with the determination of the neutrino mass hierarchy as a primary physics goal. It is also capable of observing neutrinos from terrestrial and extra-terrestrial sources, including supernova burst neutrinos, diffuse supernova neutrino background, geoneutrinos, atmospheric neutrinos, solar neutrinos, as well as exotic searches such as nucleon decays, dark matter, sterile neutrinos, etc. We present the physics motivations and the anticipated performance of the JUNO detector for various proposed measurements. By detecting reactor antineutrinos from two power plants at 53-km distance, JUNO will determine the neutrino mass hierarchy at a 3-4 sigma significance with six years of running. The measurement of antineutrino spectrum will also lead to the precise determination of three out of the six oscillation parameters to an accuracy of better than 1\%. Neutrino burst from a typical core-collapse supernova at 10 kpc would lead to ~5000 inverse-beta-decay events and ~2000 all-flavor neutrino-proton elastic scattering events in JUNO. Detection of DSNB would provide valuable information on the cosmic star-formation rate and the average core-collapsed neutrino energy spectrum. Geo-neutrinos can be detected in JUNO with a rate of ~400 events per year, significantly improving the statistics of existing geoneutrino samples. The JUNO detector is sensitive to several exotic searches, e.g. proton decay via the $p\to K^++\bar u$ decay channel. The JUNO detector will provide a unique facility to address many outstanding crucial questions in particle and astrophysics. It holds the great potential for further advancing our quest to understanding the fundamental properties of neutrinos, one of the building blocks of our Universe.

Preparation of highly photoluminescent sulfur-doped carbon dots for Fe(III) detection

Abstract: Sulfur-doped carbon dots (S-doped C-dots)were synthesized using a simple and straightforward hydrothermal method. The as-prepared S-doped C-dots exhibit significant fluorescence quantum yield (67%) and unique emission behavior. The spherical S-doped C-dots have an average diameter of 4.6 nm and the fluorescence of S-doped C-dots can be effectively and selectively quenched by Fe3+ ions. Thus, S-doped C-dots were applied as probes toward Fe3+ detection, exhibiting a limit of detection of 0.1 μM.

Conjugated Polymer–Small Molecule Alloy Leads to High Efficient Ternary Organic Solar Cells

Abstract: Ternary organic solar cells are promising candidates for bulk heterojunction solar cells; however, improving the power conversion efficiency (PCE) is quite challenging because the ternary system is complicated on phase separation behavior. In this study, a ternary organic solar cell (OSC) with two donors, including one polymer (PTB7-Th), one small molecule (p-DTS(FBTTH2)2), and one acceptor (PC71BM), is fabricated. We propose the two donors in the ternary blend forms an alloy. A notable averaged PCE of 10.5% for ternary OSC is obtained due to the improvement of the fill factor (FF) and the short-circuit current density (Jsc), and the open-circuit voltage (Voc) does not pin to the smaller Voc of the corresponding binary blends. A highly ordered face-on orientation of polymer molecules is obtained due to the formation of an alloy structure, which facilitates the enhancement of charge separation and transport and the reduction of charge recombination. This work indicates that a high crystallinity and the fac...

Significantly Enhanced Breakdown Strength and Energy Density in Sandwich-Structured Barium Titanate/Poly(vinylidene fluoride) Nanocomposites.

Abstract: Sandwich-structured BaTiO3 /poly(vinylidene fluoride) (PVDF) nanocomposites are successfully prepared by the solution-casting method layer by layer. They possess both high breakdown strength and large dielectric polarization simultaneously. An ultra-high energy-storage density of 18.8 J cm(-3) can be achieved by adjusting the volume fraction of ceramic fillers: this is almost three times larger than that of pure PVDF.

Novel biocompatible polysaccharide-based self-healing hydrogel

Abstract: A novel biocompatible polysaccharide-based self-healing hydrogel, CEC-l-OSA-l-ADH hydrogel (“l” means “linked-by”), is developed by exploiting the dynamic reaction of N-carboxyethyl chitosan (CEC) and adipic acid dihydrazide (ADH) with oxidized sodium alginate (OSA). The self-healing ability, as demonstrated by rheological recovery, macroscopic observation, and beam-shaped strain compression measurement, is attributed to the coexistence of dynamic imine and acylhydrazone bonds in the hydrogel networks. The CEC-l-OSA-l-ADH hydrogel shows excellent self-healing ability under physiological conditions with a high healing efficiency (up to 95%) without need for any external stimuli. In addition, the CEC-l-OSA-l-ADH hydrogel exhibits good cytocompatibility and cell release as demonstrated by three-dimensional cell encapsulation. With these superior properties, the developed hydrogel holds great potential for applications in various biomedical fields, e.g., as cell or drug delivery carriers.

Recent advances of the emitters for high performance deep-blue organic light-emitting diodes

Abstract: Blue organic light-emitting diodes (OLEDs) can play a critical role in the field of organic electroluminescence (EL). As the most important applications of OLEDs, both new generation full-color flat-panel displays and future energy-saving solid-state lighting sources require blue color EL to fulfill their functions properly. However, considerable challenges still exist in searching for highly efficient, color stable, and long-lifespan materials and devices that emit blue color, especially in the development of deep-blue emitters, which are indispensable for high-quality displays and lighting sources. Encouragingly, great progress has been made in the area of deep-blue OLEDs in recent years with continuous efforts made by scientists, who are responsible for the significant achievements in the field of OLEDs. Hence, in this review, the crucial tactics employed to obtain high performance deep-blue emitters are presented, including polymers, dendrimers, small organic molecules, delayed fluorescent systems, and phosphorescent emitters. Moreover, the future perspectives and ongoing challenges of this research frontier are also highlighted.

One-pot synthesis of CoFe2O4/graphene oxide hybrids and their conversion into FeCo/graphene hybrids for lightweight and highly efficient microwave absorber

Abstract: CoFe2O4/graphene oxide hybrids have been successfully fabricated via a facile one-pot polyol route, followed by chemical conversion into FeCo/graphene hybrids under H2/NH3 atmosphere. These magnetic nanocrystals were uniformly decorated on the entire graphene nanosheets without aggregation. The morphology, chemical composition and crystal structure have been characterized in detail. In particular, FeCo/graphene hybrids show significant improvement in both permeability and permittivity due to the combination of the high magnetocrystalline anisotropy of metallic FeCo and high conductivity of light-weight graphene. This leads to remarkable enhancement in microwave absorption properties. The maximum reflection loss of FeCo/graphene hybrids reaches −40.2 dB at 8.9 GHz with a matching thickness of only 2.5 mm, and the absorption bandwidth with reflection loss exceeding −10 dB is in the 3.4–18 GHz range for the absorber thickness of only 1.5–5 mm. Moreover, the experimental relationship between matching thickness and frequency is found to obey the quarter-wavelength matching model, facilitating the design of FeCo/graphene hybrid film for practical application. The results suggest that the FeCo/graphene hybrids developed here can serve as an ideal candidate for the manufacture of light-weight and high-efficiency microwave-absorbing devices.

Precision Measurement of the Helium Flux in Primary Cosmic Rays of Rigidities 1.9 GV to 3 TV with the Alpha Magnetic Spectrometer on the International Space Station

Abstract: Knowledge of the precise rigidity dependence of the helium flux is important in understanding the origin, acceleration, and propagation of cosmic rays. A precise measurement of the helium flux in primary cosmic rays with rigidity (momentum/charge) from 1.9 GV to 3 TV based on 50 million events is presented and compared to the proton flux. The detailed variation with rigidity of the helium flux spectral index is presented for the first time. The spectral index progressively hardens at rigidities larger than 100 GV. The rigidity dependence of the helium flux spectral index is similar to that of the proton spectral index though the magnitudes are different. Remarkably, the spectral index of the proton to helium flux ratio increases with rigidity up to 45 GV and then becomes constant; the flux ratio above 45 GV is well described by a single power law.

A Maximum Efficiency Point Tracking Control Scheme for Wireless Power Transfer Systems Using Magnetic Resonant Coupling

Abstract: With a good balance between power transfer distance and efficiency, wireless power transfer (WPT) using magnetic resonant coupling is preferred in many applications. Generally, WPT systems are desired to provide constant output voltage with the highest possible efficiency as power supplies. However, the highest efficiency is not achieved by the reported closed-loop WPT systems that maintain constant output voltage against coupling and load variations. In this paper, an efficiency evaluation method is put forward to evaluate the closed-loop control schemes. Furthermore, a maximum efficiency point tracking control scheme is proposed to maximize the system efficiency while regulating the output voltage. This control scheme is unique and prominent in that it fixes the operating frequency at the receiving-side resonant frequency and converts both the input voltage and the load resistance at the same time. Thus, the maximum efficiency point on the constant output voltage trajectory can be tracked dynamically. Therefore the system's output voltage can be maintained constant and its efficiency is always the highest. The experimental results show that the maximum efficiency point is tracked and a very high overall efficiency is achieved over wide ranges of coupling coefficient and load resistance.

Self-paced curriculum learning

Abstract: Curriculum learning (CL) or self-paced learning (SPL) represents a recently proposed learning regime inspired by the learning process of humans and animals that gradually proceeds from easy to more complex samples in training. The two methods share a similar conceptual learning paradigm, but differ in specific learning schemes. In CL, the curriculum is predetermined by prior knowledge, and remain fixed thereafter. Therefore, this type of method heavily relies on the quality of prior knowledge while ignoring feedback about the learner. In SPL, the curriculum is dynamically determined to adjust to the learning pace of the leaner. However, SPL is unable to deal with prior knowledge, rendering it prone to overfitting. In this paper, we discover the missing link between CL and SPL, and propose a unified framework named self-paced curriculum leaning (SPCL). SPCL is formulated as a concise optimization problem that takes into account both prior knowledge known before training and the learning progress during training. In comparison to human education, SPCL is analogous to "instructor-student-collaborative" learning mode, as opposed to "instructor-driven" in CL or "student-driven" in SPL. Empirically, we show that the advantage of SPCL on two tasks.

Design Principles for Heteroatom‐Doped Carbon Nanomaterials as Highly Efficient Catalysts for Fuel Cells and Metal–Air Batteries

Abstract: Oxygen reduction reaction/oxygen evolution reaction (ORR/OER) catalytic activities of p-orbital heteroatom-doped carbon nanomaterials are demonstrated to correlate to the combination of the electron affinity and electronegativity of doping elements, which serves as an activity descriptor for the entire family of p-block element dopants. Such a descriptor has predictive power and enables effective design of new bifunctional catalysts with enhanced ORR/OER activities.

Relaxor Ferroelectric BaTiO3–Bi(Mg2/3Nb1/3)O3 Ceramics for Energy Storage Application

Abstract: Perovskite solid solution ceramics of (1 − x)BaTiO3–xBi(Mg2/3Nb1/3)O3 (BT–BMN) (x = 0.05–0.2) were synthesized by solid-state reaction technique. The results show that the BMN addition could lower the sintering temperature of BT-based ceramics. X-ray diffraction results reveal a pure perovskite structure for all studied samples. Dielectric measurements exhibit a relaxor-like characteristic for the BT–BMN ceramics, where broadened phase transition peaks change to a temperature-stable permittivity plateau (from −50°C to 300°C) with increasing the BMN content (x = 0.2), and slim polarization–electric field hysteresis loops were observed in samples with x ≥ 0.1. The dielectric breakdown strength and electrical resistivity of BT–BMN ceramics show their maxima of 287.7 kV/cm and 1.53 × 1013 Ω cm at x = 0.15, and an energy density of about 1.13 J/cm3 is achieved in the sample of x = 0.1.

An Improved Exponential Model for Predicting Remaining Useful Life of Rolling Element Bearings

Abstract: The remaining useful life (RUL) prediction of rolling element bearings has attracted substantial attention recently due to its importance for the bearing health management. The exponential model is one of the most widely used methods for RUL prediction of rolling element bearings. However, two shortcomings exist in the exponential model: 1) the first predicting time (FPT) is selected subjectively; and 2) random errors of the stochastic process decrease the prediction accuracy. To deal with these two shortcomings, an improved exponential model is proposed in this paper. In the improved model, an adaptive FPT selection approach is established based on the $3\sigma$ interval, and particle filtering is utilized to reduce random errors of the stochastic process. In order to demonstrate the effectiveness of the improved model, a simulation and four tests of bearing degradation processes are utilized for the RUL prediction. The results show that the improved model is able to select an appropriate FPT and reduce random errors of the stochastic process. Consequently, it performs better in the RUL prediction of rolling element bearings than the original exponential model.

Co2+/Co3+ ratio dependence of electromagnetic wave absorption in hierarchical NiCo2O4–CoNiO2 hybrids

Abstract: Amorphous hierarchical NiCo2O4–CoNiO2 hybrids have been successfully fabricated via a facile one-pot hydrothermal route, followed by morphological conversion into urchin-like structured NiCo2O4–CoNiO2 nanorods and irregular-shaped hierarchical NiCo2O4–CoNiO2 polyhedral nanocrystals through air-annealing treatment at 450 °C and 650 °C, respectively. The phase structure, morphology and chemical composition have been characterized in detail. Calcined hierarchical NiCo2O4–CoNiO2 hybrids show improved microwave absorption properties, which are ascribed to the synergistic effect of dielectric CoNiO2 and NiCo2O4 phases. In particular, the calcined hierarchical NiCo2O4–CoNiO2 hybrids at 450 °C exhibit significant enhancement in complex permittivity with respect to others due to their remarkable dipole polarization and interfacial polarization. The maximum reflection loss (RL) of the calcined hierarchical NiCo2O4–CoNiO2 hybrids at 450 °C reaches −42.13 dB at 11.84 GHz with a matching thickness of 1.55 mm, and a relatively broad absorption bandwidth (RL ≤ −10 dB) in the 13.12–17.04 GHz range. Very interestingly, the electromagnetic (EM) wave absorption performance of the hierarchical NiCo2O4–CoNiO2 hybrids shows dependence on the Co2+/Co3+ ratio. The calcined NiCo2O4–CoNiO2 hybrids at 450 °C of the most defect concentration possess the best EM wave absorption ability among all the samples. The results suggest that appropriate interactions between the building blocks in hybrids can guide us to design and fabricate highly efficient EM wave absorption materials.

3D Printable Graphene Composite.

Abstract: In human being’s history, both the Iron Age and Silicon Age thrived after a matured massive processing technology was developed. Graphene is the most recent superior material which could potentially initialize another new material Age. However, while being exploited to its full extent, conventional processing methods fail to provide a link to today’s personalization tide. New technology should be ushered in. Three-dimensional (3D) printing fills the missing linkage between graphene materials and the digital mainstream. Their alliance could generate additional stream to push the graphene revolution into a new phase. Here we demonstrate for the first time, a graphene composite, with a graphene loading up to 5.6 wt%, can be 3D printable into computer-designed models. The composite’s linear thermal coefficient is below 75 ppm·°C−1 from room temperature to its glass transition temperature (Tg), which is crucial to build minute thermal stress during the printing process.

High-efficiency non-fullerene organic solar cells enabled by a difluorobenzothiadiazole-based donor polymer combined with a properly matched small molecule acceptor

Abstract: Here we report high-performance small molecule acceptor (SMA)-based organic solar cells (OSCs) enabled by the combination of a difluorobenzothiadiazole donor polymer named PffBT4T-2DT and a SMA named SF-PDI2. It is found that SF-PDI2 matches particularly well with PffBT4T-2DT and non-fullerene OSCs with an impressive VOC of 0.98 V, and a high power conversion efficiency of 6.3% is achieved. Our study shows that PffBT4T-2DT is a promising donor material for SMA-based OSCs, and the selection of a matching SMA is also important to achieve the best OSC performance.

Global Monsoon Dynamics and Climate Change

Abstract: This article provides a comprehensive review of the global monsoon that encompasses findings from studies of both modern monsoons and paleomonsoons. We introduce a definition for the global monsoon that incorporates its three-dimensional distribution and ultimate causes, emphasizing the direct drive of seasonal pressure system changes on monsoon circulation and depicting the intensity in terms of both circulation and precipitation. We explore the global monsoon climate changes across a wide range of timescales from tectonic to intraseasonal. Common features of the global monsoon are global homogeneity, regional diversity, seasonality, quasi-periodicity, irregularity, instability, and asynchroneity. We emphasize the importance of solar insolation, Earth orbital parameters, underlying surface properties, and land-air-sea interactions for global monsoon dynamics. We discuss the primary driving force of monsoon variability on each timescale and the relationships among dynamics on multiple timescales. Natural ...