arXiv subjects

Xiang Li

Publications and source records attributed to Xiang Li.

At least 1,351 records · Page 75Linked to original sources

arXiv2018

Enhanced Radar Imaging Using a Complex-valued Convolutional Neural Network

Convolutional neural networks (CNN) have been successfully employed to tackle several remote sensing tasks such as image classification and show better performance than previous techniques. For the radar imaging community, a natural question is: Can CNN be introduced to radar imaging and enhance its performance? The presented letter gives an affirmative answer to this question. We firstly propose a processing framework by which a complex-valued CNN (CV-CNN) is used to enhance radar imaging. Then we introduce two modifications to the CV-CNN to adapt it to radar imaging tasks. Subsequently, the method to generate training data is shown and some implementation details are presented. Finally, simulations and experiments are carried out, and both results show the superiority of the proposed method on imaging quality and computational efficiency.

eess.SP↗

arXiv2018

Controlling spectral energies of all harmonics in programmable way using time-domain digital coding metasurface

Modern wireless communication is one of the most important information technologies, but its system architecture has been unchanged for many years. Here, we propose a much simpler architecture for wireless communication systems based on metasurface. We firstly propose a time-domain digital coding metasurface to reach a simple but efficient method to manipulate spectral distributions of harmonics. Under dynamic modulations of phases on surface reflectivity, we could achieve accurate controls to different harmonics in a programmable way to reach many unusual functions like frequency cloaking and velocity illusion, owing to the temporal gradient introduced by digital signals encoded by '0' and '1' sequences. A theoretical model is presented and experimentally validated to reveal the nonlinear process. Based on the time-domain digital coding metasurface, we propose and realize a new wireless communication system in binary frequency-shift keying (BFSK) frame, which has much more simplified architecture than the traditional BFSK with excellent performance for real-time message transmission. The presented work, from new concept to new system, will find important applications in modern information technologies.

physics.app-ph↗

arXiv2018

Shape Robust Text Detection with Progressive Scale Expansion Network

The challenges of shape robust text detection lie in two aspects: 1) most existing quadrangular bounding box based detectors are difficult to locate texts with arbitrary shapes, which are hard to be enclosed perfectly in a rectangle; 2) most pixel-wise segmentation-based detectors may not separate the text instances that are very close to each other. To address these problems, we propose a novel Progressive Scale Expansion Network (PSENet), designed as a segmentation-based detector with multiple predictions for each text instance. These predictions correspond to different `kernels' produced by shrinking the original text instance into various scales. Consequently, the final detection can be conducted through our progressive scale expansion algorithm which gradually expands the kernels with minimal scales to the text instances with maximal and complete shapes. Due to the fact that there are large geometrical margins among these minimal kernels, our method is effective to distinguish the adjacent text instances and is robust to arbitrary shapes. The state-of-the-art results on ICDAR 2015 and ICDAR 2017 MLT benchmarks further confirm the great effectiveness of PSENet. Notably, PSENet outperforms the previous best record by absolute 6.37\% on the curve text dataset SCUT-CTW1500. Code will be available in https://github.com/whai362/PSENet.

cs.CV↗

arXiv2018

Perceive Your Users in Depth: Learning Universal User Representations from Multiple E-commerce Tasks

Tasks such as search and recommendation have become increas- ingly important for E-commerce to deal with the information over- load problem. To meet the diverse needs of di erent users, person- alization plays an important role. In many large portals such as Taobao and Amazon, there are a bunch of di erent types of search and recommendation tasks operating simultaneously for person- alization. However, most of current techniques address each task separately. This is suboptimal as no information about users shared across di erent tasks. In this work, we propose to learn universal user representations across multiple tasks for more e ective personalization. In partic- ular, user behavior sequences (e.g., click, bookmark or purchase of products) are modeled by LSTM and attention mechanism by integrating all the corresponding content, behavior and temporal information. User representations are shared and learned in an end-to-end setting across multiple tasks. Bene ting from better information utilization of multiple tasks, the user representations are more e ective to re ect their interests and are more general to be transferred to new tasks. We refer this work as Deep User Perception Network (DUPN) and conduct an extensive set of o ine and online experiments. Across all tested ve di erent tasks, our DUPN consistently achieves better results by giving more e ective user representations. Moreover, we deploy DUPN in large scale operational tasks in Taobao. Detailed implementations, e.g., incre- mental model updating, are also provided to address the practical issues for the real world applications.

stat.ML↗

arXiv2018

Probabilistic Embedding of Knowledge Graphs with Box Lattice Measures

Embedding methods which enforce a partial order or lattice structure over the concept space, such as Order Embeddings (OE) (Vendrov et al., 2016), are a natural way to model transitive relational data (e.g. entailment graphs). However, OE learns a deterministic knowledge base, limiting expressiveness of queries and the ability to use uncertainty for both prediction and learning (e.g. learning from expectations). Probabilistic extensions of OE (Lai and Hockenmaier, 2017) have provided the ability to somewhat calibrate these denotational probabilities while retaining the consistency and inductive bias of ordered models, but lack the ability to model the negative correlations found in real-world knowledge. In this work we show that a broad class of models that assign probability measures to OE can never capture negative correlation, which motivates our construction of a novel box lattice and accompanying probability measure to capture anticorrelation and even disjoint concepts, while still providing the benefits of probabilistic modeling, such as the ability to perform rich joint and conditional queries over arbitrary sets of concepts, and both learning from and predicting calibrated uncertainty. We show improvements over previous approaches in modeling the Flickr and WordNet entailment graphs, and investigate the power of the model.

stat.ML↗

arXiv2018

Search for gamma-ray emission from the nearby dwarf spheroidal galaxies with 9 years of Fermi-LAT data

In this work, we search for $γ$-ray emission from the directions of some nearby Milky Way dwarf spheroidal galaxies (dSphs) and candidates with the publicly-available Pass 8 data of Fermi-LAT. Our sample includes 12 sources with the distances $<50$ kpc. Very weak $γ$-ray excesses ($\sim 2σ$) are found in some dSphs/candidates, consistent with that reported in the previous literature. Intriguingly, the peak test statistic (TS) value of the weak emission from the direction of Reticulum II rises continually. If interpreted as dark matter (DM) annihilation, the peak TS value is 13.5 for the annihilation channel of $χχ\rightarrow τ^{+}τ^{-}$ and the DM mass of $m_χ\sim 16$ GeV. The combination of all these nearby sources yields a more significant (with local significance $> 4σ$) $γ$-ray signal.

astro-ph.HE↗

arXiv2018

Giant planets around FGK stars form probably through core accretion

We present a statistical study of the planet-metallicity (P-M) correlation, by comparing the 744 stars with candidate planets (SWPs) in the Kepler field which have been observed with LAMOST, and a sample of distance-independent, fake "twin" stars in the Kepler field with no planet reported (CKSNPs) yet. With the well-defined and carefully-selected large samples, we find for the first time a turn-off P-M correlation of Delta [Fe/H]_(SWPs-SNPs), which in average increases from ~0.00+-0.03 dex to 0.06+-0.03 dex, and to 0.12+-0.03 for stars with Earth, Neptune, Jupiter-sized planets successively, and then declines to ~-0.01+-0.03 dex for more massive planets or brown dwarfs. Moreover, the percentage of those systems with positive Delta[Fe/H] has the same turn-off pattern. We also find FG-type stars follow this general trend, but K-type stars are different. Moderate metal enhancement (~0.1-0.2 dex) for K-type stars with planets of radii between 2 to 4 Earth radius as compared to CKSNPs is observed, which indicates much higher metallicities are required for Super-Earths, Neptune-sized planets to form around K-type stars. We point out that the P-M correlation is actually metallicity-dependent, i.e., the correlation is positive at solar and super-solar metallicities, and negative at subsolar metallicities. No steady increase of Delta[Fe/H] against planet sizes is observed for rocky planets, excluding the pollution scenario as a major mechanism for the P-M correlation. All these clues suggest that giant planets probably form differently from rocky planets or more massive planets/brown dwarfs, and the core-accretion scenario is highly favoured, and high metallicity is a prerequisite for massive planets to form.

astro-ph.EP↗

arXiv2018

Field-Free Switching of Perpendicular Magnetic Tunnel Junction via Voltage-Gated Spin Hall Effect for Low-Power Spintronic Memory

Spin Hall effect (SHE) and voltage-controlled magnetic anisotropy (VCMA) are two promising methods for low-power electrical manipulation of magnetization. Recently, magnetic field-free switching of perpendicular magnetization through SHE has been reported with the aid of an exchange bias from an antiferromagnetic IrMn layer. In this letter, we experimentally demonstrate that the IrMn/CoFeB/MgO structure exhibits a VCMA effect of 39 fJ/Vm, which is comparable to that of the Ta/CoFeB/MgO structure. Magnetization dynamics under a combination of the SHE and VCMA are modeled and simulated. It is found that, by applying a voltage of 1.5 V, the critical SHE switching current can be decreased by 10 times owing to the VCMA effect, leading to low-power operations. Furthermore, a high-density spintronic memory structure can be built with multiple magnetic tunnel junctions (MTJs) located on a single IrMn strip. Through hybrid CMOS/MTJ simulations, we demonstrate that fast-speed write operations can be achieved with power consumption of only 8.5 fJ/bit. These findings reveal the possibility to realize high-density and low-power spintronic memory manipulated by voltage-gated SHE.

physics.app-ph↗

arXiv2018

Total gravitational mass of the Galactic Double Neutron Star systems: evidence for a bimodal distribution

So far, in total 15 double neutron star systems (DNSs) with a reliable measurement of the total gravitational mass ($M_{\rm T}$) have been detected in the Galaxy. In this work we study the distribution of $M_{\rm T}$. The data prefer the double Gaussian distribution over a single Gaussian distribution and the low and high mass populations center at $M_{\rm T}\sim 2.58M_\odot$ and $\sim 2.72M_\odot$, respectively. The progenitor stars of GW170817 have a $M_{\rm T}=2.74^{+0.04}_{-0.01}M_\odot$, falling into the high mass population. With a local neutron star merger rate of $\sim 10^{3}~{\rm Gpc^{-3}~yr^{-1}}$, supposing the $M_{\rm T}$ of those merging neutron stars also follow the double Gaussian distribution, the upcoming runs of the advanced LIGO/Virgo will soon detect some events with a $M_{\rm T}\lesssim 2.6M_\odot$ that can effectively probe the equation of state of the neutron stars and the distribution function is expected to be reliably reconstructed in the next decade.

astro-ph.HE↗

arXiv2018

Short GRBs: opening angles, local neutron star merger rate and off-axis events for GRB/GW association

The jet breaks in the afterglow lightcurves of short gamma-ray bursts (SGRBs), rarely detected so far, are crucial for estimating the half-opening angles of the ejecta ($θ_{\rm j}$) and hence the neutron star merger rate. In this work we report the detection of jet decline behaviors in GRB 150424A and GRB 160821B and find $θ_{\rm j}\sim 0.1$ rad. Together with five events reported before 2015 and other three "identified" recently (GRB 050709, GRB 060614 and GRB 140903A), we have a sample consisting of nine SGRBs and one long-short GRB with reasonably estimated $θ_{\rm j}$. In particular, three {\it Swift} bursts in the sample have redshifts $z\leq 0.2$, with which we estimate the local neutron star merger rate density {to be $\sim 1109^{+1432}_{-657}~{\rm Gpc^{-3}~yr^{-1}}$ or $162^{+140}_{-83} {\rm Gpc^{-3}yr^{-1}}$ if the narrowly-beamed GRB 061201 is excluded}. Inspired by the typical $θ_{\rm j}\sim 0.1$ rad found currently, we further investigate whether the off-beam GRBs (in the uniform jet model) or the off-axis events (in the structured jet model) can significantly enhance the GRB/GW association or not. For the former the enhancement is at most moderate, while for the latter the enhancement can be much greater and a high GRB/GW association probability of $\sim 10\%$ is possible. We also show that the data of GRB 160821B may contain a macronova/kilonova emission component with a temperature of $\sim 3100$ K at $\sim 3.6$ days after the burst and more data are needed to ultimately clarify.

astro-ph.HE↗

arXiv2018

A joint decomposition method for global optimization of multiscenario nonconvex mixed-integer nonlinear programs

This paper proposes a joint decomposition method that combines La- grangian decomposition and generalized Benders decomposition, to efficiently solve multiscenario nonconvex mixed-integer nonlinear programming (MINLP) problems to global optimality, without the need for explicit branch and bound search. In this approach, we view the variables coupling the scenario dependent variables and those causing nonconvexity as complicating variables. We systemat- ically solve the Lagrangian decomposition subproblems and the generalized Ben- ders decomposition subproblems in a unified framework. The method requires the solution of a difficult relaxed master problem, but the problem is only solved when necessary. Enhancements to the method are made to reduce the number of the relaxed master problems to be solved and ease the solution of each relaxed master problem. We consider two scenario-based, two-stage stochastic nonconvex MINLP problems that arise from integrated design and operation of process net- works in the case study, and we show that the proposed method can solve the two problems significantly faster than state-of-the-art global optimization solvers.

math.OC↗

arXiv2018

A new 2-D Model to analyze uncertainty sources of sparse sea surface CO2 partial pressure

In order to better comprehend the global carbon cycle and predict the prognosis for the response to climate change, accurate assessment of sea-air CO2 flux is necessary. Comparing to the relative homogeneously distribution of atmospheric CO2 , the pCO2 in the sea surface water is exposed to huge spatio-temporal variability, which leaves a prominent uncertainty resource. Many regional studies typically divided the observational pCO2 data into grid boxes so as to obtain enough data points statistically for their calculatio. However, using the data inside the grid box areas to represent its holistic property (such as standard deviation to represent spatial variance) will mix up three different uncertainty sources. First, the analytical error in the pCO2 determination and the associated environmental parameters used in deriving pCO2 . Second, the spatial variance because of inhomogenous spatial pattern of sea surface pCO2 , especially the region with a dramatically dynamic circumstance like: coastal areas, boundaries or fronts and etc. Third, the estimation process in undersampling condition, specifically, this kind of uncertainty origins from the process that using a sparse data to represent its holistic property of the box area. Common uncertainty quantification by Standard Deviation will mix up the different sources of uncertainty. In this paper, it introduces an optimized procedure to determine three sources of uncertainty (1st analytical error, 2nd spatial variance, 3rd bias from undersampling.) using the combined remote sensing-derived and field-measured pCO2 data. In order to provide a comprehensive error assessment report.

physics.ao-ph↗

arXiv2018

Adversarial Metric Learning

In the past decades, intensive efforts have been put to design various loss functions and metric forms for metric learning problem. These improvements have shown promising results when the test data is similar to the training data. However, the trained models often fail to produce reliable distances on the ambiguous test pairs due to the distribution bias between training set and test set. To address this problem, the Adversarial Metric Learning (AML) is proposed in this paper, which automatically generates adversarial pairs to remedy the distribution bias and facilitate robust metric learning. Specifically, AML consists of two adversarial stages, i.e. confusion and distinguishment. In confusion stage, the ambiguous but critical adversarial data pairs are adaptively generated to mislead the learned metric. In distinguishment stage, a metric is exhaustively learned to try its best to distinguish both the adversarial pairs and the original training pairs. Thanks to the challenges posed by the confusion stage in such competing process, the AML model is able to grasp plentiful difficult knowledge that has not been contained by the original training pairs, so the discriminability of AML can be significantly improved. The entire model is formulated into optimization framework, of which the global convergence is theoretically proved. The experimental results on toy data and practical datasets clearly demonstrate the superiority of AML to the representative state-of-the-art metric learning methodologies.

cs.LG↗

arXiv2018

Mixed Link Networks

Basing on the analysis by revealing the equivalence of modern networks, we find that both ResNet and DenseNet are essentially derived from the same "dense topology", yet they only differ in the form of connection -- addition (dubbed "inner link") vs. concatenation (dubbed "outer link"). However, both two forms of connections have the superiority and insufficiency. To combine their advantages and avoid certain limitations on representation learning, we present a highly efficient and modularized Mixed Link Network (MixNet) which is equipped with flexible inner link and outer link modules. Consequently, ResNet, DenseNet and Dual Path Network (DPN) can be regarded as a special case of MixNet, respectively. Furthermore, we demonstrate that MixNets can achieve superior efficiency in parameter over the state-of-the-art architectures on many competitive datasets like CIFAR-10/100, SVHN and ImageNet.

cs.LG↗

arXiv2018

Multiplex Influence Maximization in Online Social Networks with Heterogeneous Diffusion Models

Motivated by online social networks that are linked together through overlapping users, we study the influence maximization problem on a multiplex, with each layer endowed with its own model of influence diffusion. This problem is a novel version of the influence maximization problem that necessitates new analysis incorporating the type of propagation on each layer of the multiplex. We identify a new property, generalized deterministic submodular, which when satisfied by the propagation in each layer, ensures that the propagation on the multiplex overall is submodular -- for this case, we formulate ISF, the greedy algorithm with approximation ratio $(1 - 1/e)$. Since the size of a multiplex comprising multiple OSNs may encompass billions of users, we formulate an algorithm KSN that runs on each layer of the multiplex in parallel. KSN takes an $α$-approximation algorithm A for the influence maximization problem on a single-layer network as input, and has approximation ratio $\frac{(1-ε)α}{(o+1)k}$ for arbitrary $ε> 0$, $o$ is the number of overlapping users, and $k$ is the number of layers in the multiplex. Experiments on real and synthesized multiplexes validate the efficacy of the proposed algorithms for the problem of influence maximization in the heterogeneous multiplex. Implementations of ISF and KSN are available at http://www.alankuhnle.com/papers/mim/mim.html.

cs.SI↗

arXiv2018

Densely Connected Bidirectional LSTM with Applications to Sentence Classification

Deep neural networks have recently been shown to achieve highly competitive performance in many computer vision tasks due to their abilities of exploring in a much larger hypothesis space. However, since most deep architectures like stacked RNNs tend to suffer from the vanishing-gradient and overfitting problems, their effects are still understudied in many NLP tasks. Inspired by this, we propose a novel multi-layer RNN model called densely connected bidirectional long short-term memory (DC-Bi-LSTM) in this paper, which essentially represents each layer by the concatenation of its hidden state and all preceding layers' hidden states, followed by recursively passing each layer's representation to all subsequent layers. We evaluate our proposed model on five benchmark datasets of sentence classification. DC-Bi-LSTM with depth up to 20 can be successfully trained and obtain significant improvements over the traditional Bi-LSTM with the same or even less parameters. Moreover, our model has promising performance compared with the state-of-the-art approaches.

cs.CL↗

arXiv2018

SESR: Single Image Super Resolution with Recursive Squeeze and Excitation Networks

Single image super resolution is a very important computer vision task, with a wide range of applications. In recent years, the depth of the super-resolution model has been constantly increasing, but with a small increase in performance, it has brought a huge amount of computation and memory consumption. In this work, in order to make the super resolution models more effective, we proposed a novel single image super resolution method via recursive squeeze and excitation networks (SESR). By introducing the squeeze and excitation module, our SESR can model the interdependencies and relationships between channels and that makes our model more efficiency. In addition, the recursive structure and progressive reconstruction method in our model minimized the layers and parameters and enabled SESR to simultaneously train multi-scale super resolution in a single model. After evaluating on four benchmark test sets, our model is proved to be above the state-of-the-art methods in terms of speed and accuracy.

cs.CV↗

arXiv2018

Detection of the Characteristic Pion-Decay Signature in the Molecular Clouds

The $γ$-ray emission from molecular clouds is widely believed to have a hadronic origin, while unequivocal evidence is still lacking. In this work, we analyze the Fermi-LAT Pass 8 publicly available data accumulated from 2008 August 4 to 2017 June 29 and report the significant detection of a characteristic $π^0$-decay feature in the $γ$-ray spectra of the molecular clouds Orion A and Orion B. This detection provides a direct evidence for the hadronic origin of their $γ$-ray emission.

astro-ph.HE↗