arXiv subjects

Yan Xu

Publications and source records attributed to Yan Xu.

At least 289 records · Page 16Linked to original sources

arXiv2017

Mapping the Milky Way with LAMOST I: Method and overview

We present a statistical method to derive the stellar density profiles of the Milky Way from spectroscopic survey data, taking into account selection effects. We assume that the selection function of the spectroscopic survey is based on photometric colors and magnitudes and possibly altered during observations and data reductions. Then the underlying selection function for a line-of-sight can be well recovered by comparing the distribution of the spectroscopic stars in a color-magnitude plane with that of the photometric dataset. Subsequently, the stellar density profile along a line-of-sight can be derived from the spectroscopically measured stellar density profile multiplied by the selection function. The method is validated using Galaxia mock data with two different selection functions. We demonstrate that the derived stellar density profiles well reconstruct the true ones not only for the full targets, but also for the sub-populations selected from the full dataset. Finally, the method is applied to map the density profiles for the Galactic disk and halo, respectively, using the LAMOST RGB stars. The Galactic disk extends to about R=19 kpc, where the disk still contributes about 10% to the total stellar surface density. Beyond this radius, the disk smoothly transitions to the halo without any truncation, bending, or broken. Moreover, no over-density corresponding to the Monoceros ring is found in the Galactic anti-center direction. The disk shows moderate north-south asymmetry at radii larger than 12 kpc. On the other hand, the R-Z tomographic map directly shows that the stellar halo is substantially oblate within a Galactocentric radius of 20 kpc and gradually becomes nearly spherical beyond 30 kpc.

astro-ph.GA↗

arXiv2017

Flare differentially rotates sunspot on Sun's surface

Sunspots are concentrations of magnetic field visible on the solar surface (photosphere). It was considered implausible that solar flares, as resulted from magnetic reconnection in the tenuous corona, would cause a direct perturbation of the dense photosphere involving bulk motion. Here we report the sudden flare-induced rotation of a sunspot using the unprecedented spatiotemporal resolution of the 1.6 m New Solar Telescope, supplemented by magnetic data from the Solar Dynamics Observatory. It is clearly observed that the rotation is non-uniform over the sunspot: as the flare ribbon sweeps across, its different portions accelerate (up to 50 deg per hr) at different times corresponding to peaks of flare hard X-ray emission. The rotation may be driven by the surface Lorentz-force change due to the back reaction of coronal magnetic restructuring and is accompanied by a downward Poynting flux. These results have direct consequences for our understanding of energy and momentum transportation in the flare-related phenomena.

astro-ph.SR↗

arXiv2017

High-resolution observations of flare precursors in the low solar atmosphere

Solar flares are generally believed to be powered by free magnetic energy stored in the corona, but the build up of coronal energy alone may be insufficient for the imminent flare occurrence. The flare onset mechanism is a critical but less understood problem, insights into which could be gained from small-scale energy releases known as precursors, which are observed as small pre-flare brightenings in various wavelengths, and also from certain small-scale magnetic configurations such as the opposite polarity fluxes, where magnetic orientation of small bipoles is opposite to that of the ambient main polarities. However, high-resolution observations of flare precursors together with the associated photospheric magnetic field dynamics are lacking. Here we study precursors of a flare using unprecedented spatiotemporal resolution of the 1.6 m New Solar Telescope, complemented by novel microwave data. Two episodes of precursor brightenings are initiated at a small-scale magnetic channel (a form of opposite polarity fluxes) with multiple polarity inversions and enhanced magnetic fluxes and currents, lying near the footpoints of sheared magnetic loops. The low-atmospheric origin of these precursor emissions is corroborated by microwave spectra. We propose that the emerging magnetic channel field interacts with the sheared arcades to cause precursor brightenings at the main flare core region. These high-resolution results provide evidence of low-atmospheric small-scale energy release and possible relationship to the onset of the main flare.

astro-ph.SR↗

arXiv2017

Global Energetics of Solar Flares: V. Energy Closure in Flares and Coronal Mass Ejections

In this study we synthesize the results of four previous studies on the global energetics of solar flares and associated coronal mass ejections (CMEs), which include magnetic, thermal, nonthermal, and CME energies in 399 solar M and X-class flare events observed during the first 3.5 years of the Solar Dynamics Observatory (SDO) mission. Our findings are: (1) The sum of the mean nonthermal energy of flare-accelerated particles ($E_{\mathrm{nt}}$), the energy of direct heating ($E_{\mathrm{dir}}$), and the energy in coronal mass ejections ($E_{\mathrm{CME}}$), which are the primary energy dissipation processes in a flare, is found to have a ratio of $(E_{\mathrm{nt}}+E_{\mathrm{dir}}+ E_{\mathrm{CME}})/E_{\mathrm{mag}} = 0.87 \pm 0.18$, compared with the dissipated magnetic free energy $E_{\mathrm{mag}}$, which confirms energy closure within the measurement uncertainties and corroborates the magnetic origin of flares and CMEs; (2) The energy partition of the dissipated magnetic free energy is: $0.51\pm0.17$ in nonthermal energy of $\ge 6$ keV electrons, $0.17\pm0.17$ in nonthermal $\ge 1$ MeV ions, $0.07\pm0.14$ in CMEs, and $0.07\pm0.17$ in direct heating; (3) The thermal energy is almost always less than the nonthermal energy, which is consistent with the thick-target model; (4) The bolometric luminosity in white-light flares is comparable with the thermal energy in soft X-rays (SXR); (5) Solar Energetic Particle (SEP) events carry a fraction $\approx 0.03$ of the CME energy, which is consistent with CME-driven shock acceleration; and (6) The warm-target model predicts a lower limit of the low-energy cutoff at $e_c \approx 6$ keV, based on the mean differential emission measure (DEM) peak temperature of $T_e=8.6$ MK during flares. This work represents the first statistical study that establishes energy closure in solar flare/CME events.

astro-ph.SR↗

arXiv2017

Constrained Deep Weak Supervision for Histopathology Image Segmentation

In this paper, we develop a new weakly-supervised learning algorithm to learn to segment cancerous regions in histopathology images. Our work is under a multiple instance learning framework (MIL) with a new formulation, deep weak supervision (DWS); we also propose an effective way to introduce constraints to our neural networks to assist the learning process. The contributions of our algorithm are threefold: (1) We build an end-to-end learning system that segments cancerous regions with fully convolutional networks (FCN) in which image-to-image weakly-supervised learning is performed. (2) We develop a deep week supervision formulation to exploit multi-scale learning under weak supervision within fully convolutional networks. (3) Constraints about positive instances are introduced in our approach to effectively explore additional weakly-supervised information that is easy to obtain and enjoys a significant boost to the learning process. The proposed algorithm, abbreviated as DWS-MIL, is easy to implement and can be trained efficiently. Our system demonstrates state-of-the-art results on large-scale histopathology image datasets and can be applied to various applications in medical imaging beyond histopathology images such as MRI, CT, and ultrasound images.

cs.CV↗

arXiv2016

Optimizing Quantiles in Preference-based Markov Decision Processes

In the Markov decision process model, policies are usually evaluated by expected cumulative rewards. As this decision criterion is not always suitable, we propose in this paper an algorithm for computing a policy optimal for the quantile criterion. Both finite and infinite horizons are considered. Finally we experimentally evaluate our approach on random MDPs and on a data center control problem.

cs.AI↗

arXiv2016

End-to-End Subtitle Detection and Recognition for Videos in East Asian Languages via CNN Ensemble with Near-Human-Level Performance

In this paper, we propose an innovative end-to-end subtitle detection and recognition system for videos in East Asian languages. Our end-to-end system consists of multiple stages. Subtitles are firstly detected by a novel image operator based on the sequence information of consecutive video frames. Then, an ensemble of Convolutional Neural Networks (CNNs) trained on synthetic data is adopted for detecting and recognizing East Asian characters. Finally, a dynamic programming approach leveraging language models is applied to constitute results of the entire body of text lines. The proposed system achieves average end-to-end accuracies of 98.2% and 98.3% on 40 videos in Simplified Chinese and 40 videos in Traditional Chinese respectively, which is a significant outperformance of other existing methods. The near-perfect accuracy of our system dramatically narrows the gap between human cognitive ability and state-of-the-art algorithms used for such a task.

cs.CV↗

arXiv2016

Improved Relation Classification by Deep Recurrent Neural Networks with Data Augmentation

Nowadays, neural networks play an important role in the task of relation classification. By designing different neural architectures, researchers have improved the performance to a large extent in comparison with traditional methods. However, existing neural networks for relation classification are usually of shallow architectures (e.g., one-layer convolutional neural networks or recurrent networks). They may fail to explore the potential representation space in different abstraction levels. In this paper, we propose deep recurrent neural networks (DRNNs) for relation classification to tackle this challenge. Further, we propose a data augmentation method by leveraging the directionality of relations. We evaluated our DRNNs on the SemEval-2010 Task~8, and achieve an F1-score of 86.1%, outperforming previous state-of-the-art recorded results.

cs.CL↗

arXiv2016

How Transferable are Neural Networks in NLP Applications?

Transfer learning is aimed to make use of valuable knowledge in a source domain to help model performance in a target domain. It is particularly important to neural networks, which are very likely to be overfitting. In some fields like image processing, many studies have shown the effectiveness of neural network-based transfer learning. For neural NLP, however, existing studies have only casually applied transfer learning, and conclusions are inconsistent. In this paper, we conduct systematic case studies and provide an illuminating picture on the transferability of neural networks in NLP.

cs.CL↗

arXiv2016

Compressing Neural Language Models by Sparse Word Representations

Neural networks are among the state-of-the-art techniques for language modeling. Existing neural language models typically map discrete words to distributed, dense vector representations. After information processing of the preceding context words by hidden layers, an output layer estimates the probability of the next word. Such approaches are time- and memory-intensive because of the large numbers of parameters for word embeddings and the output layer. In this paper, we propose to compress neural language models by sparse word representations. In the experiments, the number of parameters in our model increases very slowly with the growth of the vocabulary size, which is almost imperceptible. Moreover, our approach not only reduces the parameter space to a large extent, but also improves the performance in terms of the perplexity measure.

cs.CL↗

arXiv2016

The Energetics of White-light Flares Observed by SDO/HMI and RHESSI

White-light (WL) flares have been observed and studied more than a century since the first discovery. However, some fundamental physics behind the brilliant emission remains highly controversial. One of the important facts in addressing the flare energetics is the spatialtemporal correlation between the white-light emission and the hard X-ray radiation, presumably suggesting that the energetic electrons are the energy sources. In this study, we present a statistical analysis of 25 strong flares (?greater than or equal to M5) observed simultaneously by the Helioseismic and Magnetic Imager (HMI) on board the Solar Dynamics Observatory (SDO) and the Reuven Ramaty High Energy Solar Spectroscopic Imager (RHESSI). Among these events, WL emission was detected by SDO/HMI in 13 flares, associated with HXR emission. To quantitatively describe the strength of WL emission, equivalent area (EA) is defined as the integrated contrast enhancement over the entire flaring area. Our results show that the equivalent area is inversely proportional to the HXR power index, indicating that stronger WL emission tends to be associated with larger population of high energy electrons. However, no obvious correlation is found between WL emission and flux of non-thermal electrons at 50 keV. For the other group of 13 flares without detectable WL emission, the HXR spectra are softer (larger power index) than those flares with WL emission, especially for the X-class flares in this group.

astro-ph.SR↗

arXiv2016

Distilling Word Embeddings: An Encoding Approach

Distilling knowledge from a well-trained cumbersome network to a small one has recently become a new research topic, as lightweight neural networks with high performance are particularly in need in various resource-restricted systems. This paper addresses the problem of distilling word embeddings for NLP tasks. We propose an encoding approach to distill task-specific knowledge from a set of high-dimensional embeddings, which can reduce model complexity by a large margin as well as retain high accuracy, showing a good compromise between efficiency and performance. Experiments in two tasks reveal the phenomenon that distilling knowledge from cumbersome embeddings is better than directly training neural networks with small embeddings.

cs.CL↗

arXiv2016

Gland Instance Segmentation by Deep Multichannel Neural Networks

In this paper, we propose a new image instance segmentation method that segments individual glands (instances) in colon histology images. This is a task called instance segmentation that has recently become increasingly important. The problem is challenging since not only do the glands need to be segmented from the complex background, they are also required to be individually identified. Here we leverage the idea of image-to-image prediction in recent deep learning by building a framework that automatically exploits and fuses complex multichannel information, regional, location and boundary patterns in gland histology images. Our proposed system, deep multichannel framework, alleviates heavy feature design due to the use of convolutional neural networks and is able to meet multifarious requirement by altering channels. Compared to methods reported in the 2015 MICCAI Gland Segmentation Challenge and other currently prevalent methods of instance segmentation, we observe state-of-the-art results based on a number of evaluation metrics.

cs.CV↗

arXiv2016

Gland Instance Segmentation by Deep Multichannel Side Supervision

In this paper, we propose a new image instance segmentation method that segments individual glands (instances) in colon histology images. This is a task called instance segmentation that has recently become increasingly important. The problem is challenging since not only do the glands need to be segmented from the complex background, they are also required to be individually identified. Here we leverage the idea of image-to-image prediction in recent deep learning by building a framework that automatically exploits and fuses complex multichannel information, regional and boundary patterns, with side supervision (deep supervision on side responses) in gland histology images. Our proposed system, deep multichannel side supervision (DMCS), alleviates heavy feature design due to the use of convolutional neural networks guided by side supervision. Compared to methods reported in the 2015 MICCAI Gland Segmentation Challenge, we observe state-of-the-art results based on a number of evaluation metrics.

cs.CV↗

arXiv2016

Multi-wavelength Study of Transition Region Penumbral Subarcsecond Bright Dots Using IRIS and NST

Using high-resolution transition region (TR) observations taken by the Interface Region Imaging Spectrograph (IRIS) mission, Tian et al. (2014b) revealed numerous short-lived subarcsecond bright dots (BDs) above sunspots (mostly located in the penumbrae), which indicate yet unexplained small-scale energy releases. Moreover, whether these subarcsecond TR brightenings have any signature in the lower atmosphere and how they are formed are still not fully resolved. This paper presents a multi-wavelength study of the TR penumbral BDs using a coordinated observation of a near disk-center sunspot with IRIS and the 1.6 m New Solar Telescope (NST) at the Big Bear Solar Observatory. NST provides high-resolution chromospheric and photospheric observations with narrow-band H-alpha imaging spectroscopy and broad-band TiO images, respectively, complementary to IRIS TR observations. A total of 2692 TR penumbral BDs are identified from a 37-minute time series of IRIS 1400 A slitjaw images. Their locations tend to be associated more with downflowing and darker fibrils in the chromosphere, and weakly associated with bright penumbral features in the photosphere. However, temporal evolution analyses of the BDs show that there is no consistent and convincing brightening response in the chromosphere. These results are compatible with a formation mechanism of the TR penumbral BDs by falling plasma from coronal heights along more vertical and dense magnetic loops. The BDs may also be produced by small-scale impulsive magnetic reconnection taking place sufficiently high in the atmosphere that has no energy release in the chromosphere.

astro-ph.SR↗

arXiv2016

Natural Language Inference by Tree-Based Convolution and Heuristic Matching

In this paper, we propose the TBCNN-pair model to recognize entailment and contradiction between two sentences. In our model, a tree-based convolutional neural network (TBCNN) captures sentence-level semantics; then heuristic matching layers like concatenation, element-wise product/difference combine the information in individual sentences. Experimental results show that our model outperforms existing sentence encoding-based approaches by a large margin.

cs.CL↗

arXiv2016

Direct Urca processes involving singlet proton superfluidity in neutron star cooling

A detailed description of the baryon direct Urca processes A: $n\rightarrow p+e+\barν_{e}$, B: $Λ\rightarrow p+e+\barν_{e}$, C: $Ξ^{-}\rightarrowΛ+e+\barν_{e}$ related to the neutron star cooling is given in the relativistic mean field approximation.The contributions of the reactions B and C on the neutrino luminosity are calculated by means of the relativistic expressions of the neutrino energy losses.Our results show that the total neutrino luminosities of the reactions A, B, C within the mass range 1.603-2.067$M_{\odot}$ (1.515-1.840$M_{\odot}$ for TM1 model) for GM1 model are larger than the corresponding values for neutron stars in npe$μ$ matter. Although the hyperon direct Urca processes B and C reduce the neutrino emissivity of the reaction A, it illustrates the reactions B and C still make the total neutrino luminosity enhancement in the above mentioned areas.Furthermore, when we only consider the $^{1}S_{0}$ proton superfluidity in neutron star cooling, we find that although the neutrino emissivity of the reactions A and B is suppressed with the appearance of $^{1}S_{0}$ proton superfluidity, the total contribution of the reactions A, B, C can still quicken a massive neutron star cooling.These results could be used to help prove appearing hyperons in PSR J1614-2230 and J0348+0432 from neutron star cooling perspective.

nucl-th↗

arXiv2016

Unprecedented Fine Structure of a Solar Flare Revealed by the 1.6~m New Solar Telescope

Solar flares signify the sudden release of magnetic energy and are sources of so called space weather. The fine structures (below 500 km) of flares are rarely observed and are accessible to only a few instruments world-wide. Here we present observation of a solar flare using exceptionally high resolution images from the 1.6~m New Solar Telescope (NST) equipped with high order adaptive optics at Big Bear Solar Observatory (BBSO). The observation reveals the process of the flare in unprecedented detail, including the flare ribbon propagating across the sunspots, coronal rain (made of condensing plasma) streaming down along the post-flare loops, and the chromosphere's response to the impact of coronal rain, showing fine-scale brightenings at the footpoints of the falling plasma. Taking advantage of the resolving power of the NST, we measure the cross-sectional widths of flare ribbons, post-flare loops and footpoint brighenings, which generally lie in the range of 80-200 km, well below the resolution of most current instruments used for flare studies. Confining the scale of such fine structure provides an essential piece of information in modeling the energy transport mechanism of flares, which is an important issue in solar and plasma physics.

astro-ph.SR↗