arXiv subjects

Li Zhang

Publications and source records attributed to Li Zhang.

At least 829 records · Page 46Linked to original sources

arXiv2019

Fast Online Object Tracking and Segmentation: A Unifying Approach

In this paper we illustrate how to perform both visual object tracking and semi-supervised video object segmentation, in real-time, with a single simple approach. Our method, dubbed SiamMask, improves the offline training procedure of popular fully-convolutional Siamese approaches for object tracking by augmenting their loss with a binary segmentation task. Once trained, SiamMask solely relies on a single bounding box initialisation and operates online, producing class-agnostic object segmentation masks and rotated bounding boxes at 55 frames per second. Despite its simplicity, versatility and fast speed, our strategy allows us to establish a new state of the art among real-time trackers on VOT-2018, while at the same time demonstrating competitive performance and the best speed for the semi-supervised video object segmentation task on DAVIS-2016 and DAVIS-2017. The project website is http://www.robots.ox.ac.uk/~qwang/SiamMask.

cs.CV↗

arXiv2019

First insight into the regime transition dynamics of mode-locked fibre lasers

Mode-locked lasers exhibit complex nonlinear dynamics. Precise observation of these dynamics will aid in understanding of the underlying physics and provide new insights for laser design and applications. The starting dynamics, from initial noise fluctuations to the mode-locking regime, have previously been observed directly by time-stretched transform-based real-time spectroscopy. However, the regime transition dynamics, which are essential processes in mode-locked lasers, have not yet been resolved because regime transition process tracking is very challenging. Here we demonstrate the first insight into the regime transition dynamics enabled by our design of a real-time programmable mode-locked fibre laser, in which different operating regimes can be achieved and switched automatically. The regime transition dynamics among initial noise fluctuations, Q-switching, fundamental mode-locking and harmonic mode-locking regimes have been observed and thoroughly analysed by both temporal and spectral means. These findings will enrich our understanding of the complex dynamics inside mode-locked lasers.

physics.optics↗

arXiv2019

On the Difficulty of Evaluating Baselines: A Study on Recommender Systems

Numerical evaluations with comparisons to baselines play a central role when judging research in recommender systems. In this paper, we show that running baselines properly is difficult. We demonstrate this issue on two extensively studied datasets. First, we show that results for baselines that have been used in numerous publications over the past five years for the Movielens 10M benchmark are suboptimal. With a careful setup of a vanilla matrix factorization baseline, we are not only able to improve upon the reported results for this baseline but even outperform the reported results of any newly proposed method. Secondly, we recap the tremendous effort that was required by the community to obtain high quality results for simple methods on the Netflix Prize. Our results indicate that empirical findings in research papers are questionable unless they were obtained on standardized benchmarks where baselines have been tuned extensively by the research community.

cs.IR↗

arXiv2019

Automated Segmentation of Pulmonary Lobes using Coordination-Guided Deep Neural Networks

The identification of pulmonary lobes is of great importance in disease diagnosis and treatment. A few lung diseases have regional disorders at lobar level. Thus, an accurate segmentation of pulmonary lobes is necessary. In this work, we propose an automated segmentation of pulmonary lobes using coordination-guided deep neural networks from chest CT images. We first employ an automated lung segmentation to extract the lung area from CT image, then exploit volumetric convolutional neural network (V-net) for segmenting the pulmonary lobes. To reduce the misclassification of different lobes, we therefore adopt coordination-guided convolutional layers (CoordConvs) that generate additional feature maps of the positional information of pulmonary lobes. The proposed model is trained and evaluated on a few publicly available datasets and has achieved the state-of-the-art accuracy with a mean Dice coefficient index of 0.947 $\pm$ 0.044.

cs.CV↗

arXiv2019

Impact of the first order antiferromagnetic phase transition on the paramagnetic spin excitations and nematic phase of SrFe$_2$As$_2$

Understanding the nature of the electronic nematic phase in iron pnictide superconductors is important for elucidating its impact on high-temperature superconductivity. Here we use transport and inelastic neutron scattering to study spin excitations and in-plane resistivity anisotropy in uniaxial pressure detwinned BaFe$_2$As$_2$ and SrFe$_2$As$_2$, the parent compounds of iron pnictide superconductors. While BaFe$_2$As$_2$ exhibits weakly first order tetragonal-to-orthorhombic structural and antiferromagnetic (AF) phase transitions below $T_s > T_N\approx 138$ K, SrFe$_2$As$_2$ has strongly coupled first order structural and AF transitions below $T_s= T_N\approx 210$ K. We find that the direct signatures of the nematic phase persist to lower temperatures above the phase transition in the case of SrFe$_2$As$_2$ compared to BaFe$_2$As$_2$. Our findings support the conclusion that the strongly first-order nature of the magnetic transition in SrFe$_2$As$_2$ weakens the nematic phase and resistivity anisotropy in the system.

cond-mat.supr-con↗

arXiv2019

Multi-Label Transfer Learning for Multi-Relational Semantic Similarity

Multi-relational semantic similarity datasets define the semantic relations between two short texts in multiple ways, e.g., similarity, relatedness, and so on. Yet, all the systems to date designed to capture such relations target one relation at a time. We propose a multi-label transfer learning approach based on LSTM to make predictions for several relations simultaneously and aggregate the losses to update the parameters. This multi-label regression approach jointly learns the information provided by the multiple relations, rather than treating them as separate tasks. Not only does this approach outperform the single-task approach and the traditional multi-task learning approach, but it also achieves state-of-the-art performance on all but one relation of the Human Activity Phrase dataset.

cs.CL↗

arXiv2019

Monte Carlo Neural Fictitious Self-Play: Approach to Approximate Nash equilibrium of Imperfect-Information Games

Researchers on artificial intelligence have achieved human-level intelligence in large-scale perfect-information games, but it is still a challenge to achieve (nearly) optimal results (in other words, an approximate Nash Equilibrium) in large-scale imperfect-information games (i.e. war games, football coach or business strategies). Neural Fictitious Self Play (NFSP) is an effective algorithm for learning approximate Nash equilibrium of imperfect-information games from self-play without prior domain knowledge. However, it relies on Deep Q-Network, which is off-line and is hard to converge in online games with changing opponent strategy, so it can't approach approximate Nash equilibrium in games with large search scale and deep search depth. In this paper, we propose Monte Carlo Neural Fictitious Self Play (MC-NFSP), an algorithm combines Monte Carlo tree search with NFSP, which greatly improves the performance on large-scale zero-sum imperfect-information games. Experimentally, we demonstrate that the proposed Monte Carlo Neural Fictitious Self Play can converge to approximate Nash equilibrium in games with large-scale search depth while the Neural Fictitious Self Play can't. Furthermore, we develop Asynchronous Neural Fictitious Self Play (ANFSP). It use asynchronous and parallel architecture to collect game experience. In experiments, we show that parallel actor-learners have a further accelerated and stabilizing effect on training.

cs.AI↗

arXiv2019

Hierarchical method for cataract grading based on retinal images using improved Haar wavelet

Cataracts, which are lenticular opacities that may occur at different lens locations, are the leading cause of visual impairment worldwide. Accurate and timely diagnosis can improve the quality of life of cataract patients. In this paper, a feature extraction-based method for grading cataract severity using retinal images is proposed. To obtain more appropriate features for the automatic grading, the Haar wavelet is improved according to the characteristics of retinal images. Retinal images of non-cataract, as well as mild, moderate, and severe cataracts, are automatically recognized using the improved Haar wavelet. A hierarchical strategy is used to transform the four-class classification problem into three adjacent two-class classification problems. Three sets of two-class classifiers based on a neural network are trained individually and integrated together to establish a complete classification system. The accuracies of the two-class classification (cataract and non-cataract) and four-class classification are 94.83% and 85.98%, respectively. The performance analysis demonstrates that the improved Haar wavelet feature achieves higher accuracy than the original Haar wavelet feature, and the fusion of three sets of two-class classifiers is superior to a simple four-class classifier. The discussion indicates that the retinal image-based method offers significant potential for cataract detection.

cs.CV↗

arXiv2019

Multi-grid Multi-Level Monte Carlo Method for Stokes-Darcy interface Model with Random Hydraulic Conductivity

In this article we develop a multi-grid multi-level Monte Carlo (MGMLMC) method for the stochastic Stokes-Darcy interface model with random hydraulic conductivity both in the porous media domain and on the interface. Because the randomness through the interface affects the flow in the Stokes domain, we investigate the coupled stochastic Stokes-Darcy model to improve the fidelity as this model also considers the second and third porosity of the free flow. Then we prove the existence and uniqueness of the weak solution of the variational form. For the numerical solution, we adopt the Monte Carlo (MC) method and finite element method (FEM), for the discrete form in the probability space and physical space, respectively. In the traditional single-level Monte Carlo (SLMC) method, more accurate numerical approximate requires both larger number of samples in probability space and smaller mesh size in the physical space. Then the computational cost increase significantly as the mesh size becomes smaller for the more accurate numerical approximate. Therefore we adopt the multi-level Monte Carlo (MLMC) method to dramatically reduce the computational cost in the probability space, because the number of samples decays fast while the mesh size decreases. We also develop a strategy to calculate the number of samples needed in MLMC method for the stochastic Stokes-Darcy model. Furthermore MLMC naturally provides the hierarchial grids and sufficient information on these grids for multi-grid (MG) method, which can in turn improve the efficiency of MLMC. In order to fully make use of the dynamical interaction between this two methods, we propose the multi-grid multi-level Monte Carlo method for more efficiently solving the stochastic model. Numerical examples are provided to verify and illustrate the proposed method and the theoretical conclusions.

math.NA↗

arXiv2019

Field-aware Neural Factorization Machine for Click-Through Rate Prediction

Recommendation systems and computing advertisements have gradually entered the field of academic research from the field of commercial applications. Click-through rate prediction is one of the core research issues because the prediction accuracy affects the user experience and the revenue of merchants and platforms. Feature engineering is very important to improve click-through rate prediction. Traditional feature engineering heavily relies on people's experience, and is difficult to construct a feature combination that can describe the complex patterns implied in the data. This paper combines traditional feature combination methods and deep neural networks to automate feature combinations to improve the accuracy of click-through rate prediction. We propose a mechannism named 'Field-aware Neural Factorization Machine' (FNFM). This model can have strong second order feature interactive learning ability like Field-aware Factorization Machine, on this basis, deep neural network is used for higher-order feature combination learning. Experiments show that the model has stronger expression ability than current deep learning feature combination models like the DeepFM, DCN and NFM.

cs.LG↗

arXiv2019

Zero-energy proton dissociation of H$_2^+$ through stimulated Raman scattering

We show that (near-)zero energy proton emission from H$_2^+$ in strong two-color and broadband laser fields is dominated by a stimulated Raman scattering process taking place on the electronic ground state. It is furthermore shown that in the (near-)zero energy region the asymmetry in proton ejection induced by asymmetric laser fields is due to the interplay of several processes, rather than only pathway interferences, with vibrational trapping (or bond-hardening) taking a key role.

physics.chem-ph↗

arXiv2019

Higher-order photonic topological states in surface-wave photonic crystals

Photonic topological states have revolutionized our understanding on the propagation and scattering of light. Recent discovery of higher-order photonic topological insulators opens an emergent horizon for zero-dimensional topological corner states. However, the previous realizations of higher-order photonic topological insulators suffer from either a limited operational frequency range due to the lumped components involved or a bulky structure with a large footprint, which are unfavorable for future integrated photonics. To overcome these limitations, we hereby experimentally demonstrate a planar surface-wave photonic crystal realization of two-dimensional higher-order topological insulators. The surface-wave photonic crystals exhibit a very large bulk bandgap (a bandwidth of 28%) due to multiple Bragg scatterings and host one-dimensional gapped edge states described by massive Dirac equations. The topology of those higher-dimensional photonic bands leads to the emergence of zero-dimensional corner states, which provide a route toward robust cavity modes for scalable, integrated photonic chips and an interface for the control of light-matter interaction.

physics.app-ph↗

arXiv2019

Detailed polarization measurements of the prompt emission of five Gamma-Ray Bursts

Gamma-ray bursts are the strongest explosions in the Universe since the Big Bang, believed to be produced either in forming black holes at the end of massive star evolution or merging of compact objects. Spectral and timing properties of gamma-ray bursts suggest that the observed bright gamma-rays are produced in the most relativistic jets in the Universe; however, the physical properties, especially the structure and magnetic topologies in the jets are still not well known, despite several decades of studies. It is widely believed that precise measurements of the polarization properties of gamma-ray bursts should provide crucial information on the highly relativistic jets. As a result there have been many reports of gamma-ray burst polarization measurements with diverse results, see, however many such measurements suffered from substantial uncertainties, mostly systematic. After the first successful measurements by the GAP and COSI instruments, here we report a statistically meaningful sample of precise polarization measurements, obtained with the dedicated gamma-ray burst polarimeter, POLAR onboard China's Tiangong-2 spacelab. Our results suggest that the gamma-ray emission is at most polarized at a level lower than some popular models have predicted; although our results also show intrapulse evolution of the polarization angle. This indicates that the low polarization degrees could be due to an evolving polarization angle during a gamma-ray burst.

astro-ph.HE↗

arXiv2019

Realization of a three-dimensional photonic topological insulator

Confining photons in a finite volume is in high demand in modern photonic devices. This motivated decades ago the invention of photonic crystals, featured with a photonic bandgap forbidding light propagation in all directions. Recently, inspired by the discoveries of topological insulators (TIs), the confinement of photons with topological protection has been demonstrated in two-dimensional (2D) photonic structures known as photonic TIs, with promising applications in topological lasers and robust optical delay lines. However, a fully three-dimensional (3D) topological photonic bandgap has never before been achieved. Here, we experimentally demonstrate a 3D photonic TI with an extremely wide (> 25% bandwidth) 3D topological bandgap. The sample consists of split-ring resonators (SRRs) with strong magneto-electric coupling and behaves as a 'weak TI', or a stack of 2D quantum spin Hall insulators. Using direct field measurements, we map out both the gapped bulk bandstructure and the Dirac-like dispersion of the photonic surface states, and demonstrate robust photonic propagation along a non-planar surface. Our work extends the family of 3D TIs from fermions to bosons and paves the way for applications in topological photonic cavities, circuits, and lasers in 3D geometries.

physics.class-ph↗

arXiv2018

Valley-Hall photonic topological insulators with dual-band kink states

Extensive researches have revealed that valley, a binary degree of freedom (DOF), can be an excellent candidate of information carrier. Recently, valley DOF has been introduced into photonic systems, and several valley-Hall photonic topological insulators (PTIs) have been experimentally demonstrated. However, in the previous valley-Hall PTIs, topological kink states only work at a single frequency band, which limits potential applications in multiband waveguides, filters, communications, and so on. To overcome this challenge, here we experimentally demonstrate a valley-Hall PTI, where the topological kink states exist at two separated frequency bands, in a microwave substrate-integrated circuitry. Both the simulated and experimental results demonstrate the dual-band valley-Hall topological kink states are robust against the sharp bends of the internal domain wall with negligible inter-valley scattering. Our work may pave the way for multi-channel substrate-integrated photonic devices with high efficiency and high capacity for information communications and processing.

physics.app-ph↗

arXiv2018

Superconductivity, Electronic Phase Diagram and Pressure Effect in Sr_{1-x}Pr_{x}FBiS_{2}

Based on a combination of X-ray diffraction, electrical transports, magnetic susceptibility, specific heat and pressure effect measurements, we report a series of BiS$_{2}$-based Sr$_{1-x}$Pr$_{x}$FBiS$_{2}$ superconductors with the maximum $T_c$ of 2.7 K for $x =$ 0.5 at ambient pressure. Superconductivity can be only observed at 0.4 $\leq$ x $\leq$ 0.7 while the normal state resistivity shows the semiconducting-like behaviors. $χ(T)$ displays the low superconducting shielding volume fractions and $C(T)$ does not exhibit distinguishable anomaly around $T_c$, suggesting a filamentary superconductivity in the Pr-doped polycrystalline samples. By varying doping concentrations, an electrnoic phase diagram can thus be established. Upon applying pressure on the optimal doping Sr$_{0.5}$Pr$_{0.5}$FBiS$_{2}$ system, $T_c$ is abruptly enhanced, reaches 8.5 K at a critical pressure of $P_c =$ 1.5 GPa and remains a slight increase up to 9.7 K up to 2.5 GPa. Accompanied with the enhancement of superconductivity from the low-$T_c$ phase to the high-$T_c$ phase, the normal state undergoes a semiconductor-to-metal transition under pressure. The scenario can be linked to the enhancement of orbital overlap of Bi-6p and S-p, contributing to the bulk superconductivity above $P_c$. Phase diagram for Sr$_{0.5}$Pr$_{0.5}$FBiS$_{2}$ is also obtained.

cond-mat.supr-con↗

arXiv2018

ALMA Observations of Ethyl Formate toward Orion KL

Orion KL is one of the prime templates of astrochemical and prebiotic chemical studies. We wish to explore more organic molecules with increasing complexity in this region. In particular, we have searched for one of the most complex organic molecules detected in space so far, ethyl formate (C$_{2}$H$_{5}$OCHO). This species is the next step in chemical complexity after the simplest member of esters (methyl formate, CH$_{3}$OCHO). The mechanisms leading to its formation are still poorly known. We have used high angular resolution ($\sim$ 1.$^{\prime\prime}$5) ALMA observations covering a large bandwidth from 214 to 247 GHz. We have detected 82 unblended lines of C$_{2}$H$_{5}$OCHO (49 and 33 of the trans and gauche conformers, respectively). The line images showed that C$_{2}$H$_{5}$OCHO arises mainly from the compact ridge and the hot core-southwest regions. The derived rotational temperatures and column densities are 122 $\pm$ 34 K, (0.9 $\pm$ 0.3) $\times$ 10$^{16}$ cm$^{-2}$ for the hot core-SW, and 103 $\pm$ 13 K, (0.6 $\pm$ 0.3) $\times$ 10$^{16}$ cm$^{-2}$ for the compact ridge. The comparison of spatial distribution and abundance ratios with chemically related molecules (methyl formate, ethanol and formic acid) indicates that C$_{2}$H$_{5}$OCHO is likely formed on the surface of dust grains by addition of CH$_{3}$ to functional-group radicals (CH$_{2}$OCHO) derived from methyl formate (CH$_{3}$OCHO).

astro-ph.SR↗

arXiv2018

Automated segmentaiton and classification of arterioles and venules using Cascading Dilated Convolutional Neural Networks

The change of retinal vasculature is an early sign of many vascular and systematic diseases, such as diabetes and hypertension. Different behaviors of retinal arterioles and venules form an important metric to measure the disease severity. Therefore, an accurate classification of arterioles and venules is of great necessity. In this work, we propose a novel architecture of deep convolutional neural network for segmenting and classifying arterioles and venules on retinal fundus images. This network takes the original color fundus image as inputs and multi-class labels as outputs. We adopt the encoding-decoding structure (Unet) as the backbone network of our proposed model. To improve the classification accuracy, we develop a special encoding path that couples InceptionV4 modules and Cascading Dilated Convolutions (CDCs) on top of the backbone network. The model is thus able to extract and fuse high-level semantic features from multi-scale receptive fields. The proposed method has outperformed the previous state-of-the-art method on DRIVE dataset with an accuracy of 0.955 $\pm$ 0.002.

cs.CV↗