arXiv subjects

Dong Xu

Publications and source records attributed to Dong Xu.

At least 217 records · Page 12Linked to original sources

arXiv2017

Image Projective Invariants

In this paper, we propose relative projective differential invariants (RPDIs) which are invariant to general projective transformations. By using RPDIs and the structural frame of integral invariant, projective weighted moment invariants (PIs) can be constructed very easily. It is first proved that a kind of projective invariants exists in terms of weighted integration of images, with relative differential invariants as the weight functions. Then, some simple instances of PIs are given. In order to ensure the stability and discriminability of PIs, we discuss how to calculate partial derivatives of discrete images more accurately. Since the number of pixels in discrete images before and after the geometric transformation may be different, we design the method to normalize the number of pixels. These ways enhance the performance of PIs. Finally, we carry out some experiments based on synthetic and real image datasets. We choose commonly used moment invariants for comparison. The results indicate that PIs have better performance than other moment invariants in image retrieval and classification. With PIs, one can compare the similarity between images under the projective transformation without knowing the parameters of the transformation, which provides a good tool to shape analysis in image processing, computer vision and pattern recognition.

cs.CV↗

arXiv2017

A New Measurement of the Spectral Lag of Gamma-Ray Bursts and its Implications for Spectral Evolution Behaviors

We carry out a systematical study of the spectral lag properties of 50 single-pulsed Gamma-Ray Bursts (GRBs) detected by Fermi/GBM. By dividing the light curves into multiple consecutive energy channels we provide a new measurement of the spectral lag which is independent on energy channel selections. We perform a detailed statistical study of our new measurements. We find two similar power-law energy dependencies of both the pulse arrival time and pulse width. Our new results on the power-law indices would favor the relativistic geometric effects for the origin of spectral lag. However, a complete theoretical framework that can fully account for the diverse energy dependencies of both arrival time and pulse width revealed in this work is still missing. We also study the spectral evolution behaviors of the GRB pulses. We find that the GRB pulse with negligible spectral lag would usually have a shorter pulse duration and would appear to have a "hardness-intensity tracking" (HIT) behavior and the GRB pulse with a significant spectral lag would usually have a longer pulse duration and would appear to have a "hard-to-soft" (HTS) behavior.

astro-ph.HE↗

arXiv2017

Skeleton-Based Action Recognition Using Spatio-Temporal LSTM Network with Trust Gates

Skeleton-based human action recognition has attracted a lot of research attention during the past few years. Recent works attempted to utilize recurrent neural networks to model the temporal dependencies between the 3D positional configurations of human body joints for better analysis of human activities in the skeletal data. The proposed work extends this idea to spatial domain as well as temporal domain to better analyze the hidden sources of action-related information within the human skeleton sequences in both of these domains simultaneously. Based on the pictorial structure of Kinect's skeletal data, an effective tree-structure based traversal framework is also proposed. In order to deal with the noise in the skeletal data, a new gating mechanism within LSTM module is introduced, with which the network can learn the reliability of the sequential data and accordingly adjust the effect of the input data on the updating procedure of the long-term context representation stored in the unit's memory cell. Moreover, we introduce a novel multi-modal feature fusion strategy within the LSTM unit in this paper. The comprehensive experimental results on seven challenging benchmark datasets for human action recognition demonstrate the effectiveness of the proposed method.

cs.CV↗

arXiv2017

Shape DNA: Basic Generating Functions for Geometric Moment Invariants

Geometric moment invariants (GMIs) have been widely used as basic tool in shape analysis and information retrieval. Their structure and characteristics determine efficiency and effectiveness. Two fundamental building blocks or generating functions (GFs) for invariants are discovered, which are dot product and vector product of point vectors in Euclidean space. The primitive invariants (PIs) can be derived by carefully selecting different products of GFs and calculating the corresponding multiple integrals, which translates polynomials of coordinates of point vectors into geometric moments. Then the invariants themselves are expressed in the form of product of moments. This procedure is just like DNA encoding proteins. All GMIs available in the literature can be decomposed into linear combinations of PIs. This paper shows that Hu's seven well known GMIs in computer vision have a more deep structure, which can be further divided into combination of simpler PIs. In practical uses, low order independent GMIs are of particular interest. In this paper, a set of PIs for similarity transformation and affine transformation in 2D are presented, which are simpler to use, and some of which are newly reported. The discovery of the two generating functions provides a new perspective of better understanding shapes in 2D and 3D Euclidean spaces, and the method proposed can be further extended to higher dimensional spaces and different manifolds, such as curves, surfaces and so on.

cs.CV↗

arXiv2017

Waiting times and stopping probabilities for patterns in Markov chains

Suppose that $\mathcal C$ is a finite collection of patterns. Observe a Markov chain until one of the patterns in $\mathcal C$ occurs as a run. This time is denoted by $τ$. In this paper, we aim to give an easy way to calculate the mean waiting time $E(τ)$ and the stopping probabilities $P(τ=τ_A)$ with $A\in\mathcal C$, where $τ_A$ is the waiting time until the pattern $A$ appears as a run.

math.PR↗

arXiv2016

Solving the 56Ni puzzle of magnetar-powered broad-lined type Ic supernovae

Broad-lined type Ic supernovae (SNe Ic-BL) are of great importance because their association with long-duration gamma-ray bursts (LGRBs) holds the key to deciphering the central engine of LGRBs, which refrains from being unveiled despite decades of investigation. Among the two popularly hypothesized types of central engine, i.e., black holes and strongly magnetized neutron stars (magnetars), there is mounting evidence that the central engine of GRB-associated SNe (GRB-SNe) is rapidly rotating magnetars. Theoretical analysis also suggests that magnetars could be the central engine of SNe Ic-BL. What puzzled the researchers is the fact that light curve modeling indicates that as much as 0.2-0.5 solar mass of 56Ni was synthesized during the explosion of the SNe Ic-BL, which is unfortunately in direct conflict with current state-of-the-art understanding of magnetar-powered 56Ni synthesis. Here we propose a dynamic model of magnetar-powered SNe to take into account the acceleration of the ejecta by the magnetar, as well as the thermalization of the injected energy. Assuming that the SN kinetic energy comes exclusively from the magnetar acceleration, we find that although a major fraction of the rotational energy of the magnetar is to accelerate the SNe ejecta, a tiny fraction of this energy deposited as thermal energy of the ejecta is enough to reduce the needed 56Ni to 0.06 solar mass for both SN 1997ef and SN 2007ru. We therefore suggest that magnetars could power SNe Ic-BL both in aspects of energetics and of 56Ni synthesis.

astro-ph.HE↗

arXiv2016

Gamma-ray burst progenitors

We review our current understanding of the progenitors of both long and short duration gamma-ray bursts (GRBs). Constraints can be derived from multiple directions, and we use three distinct strands; i) direct observations of GRBs and their host galaxies, ii) parameters derived from modeling, both via population synthesis and direct numerical simulation and iii) our understanding of plausible analog progenitor systems observed in the local Universe. From these joint constraints, we describe the likely routes that can drive massive stars to the creation of long GRBs, and our best estimates of the scenarios that can create compact object binaries which will ultimately form short GRBs, as well as the associated rates of both long and short GRBs. We further discuss how different the progenitors may be in the case of black hole engine or millisecond-magnetar models for the production of GRBs, and how central engines may provide a unifying theme between many classes of extremely luminous transient, from luminous and super-luminous supernovae to long and short GRBs.

astro-ph.HE↗

arXiv2016

A Siamese Long Short-Term Memory Architecture for Human Re-Identification

Matching pedestrians across multiple camera views known as human re-identification (re-identification) is a challenging problem in visual surveillance. In the existing works concentrating on feature extraction, representations are formed locally and independent of other regions. We present a novel siamese Long Short-Term Memory (LSTM) architecture that can process image regions sequentially and enhance the discriminative capability of local feature representation by leveraging contextual information. The feedback connections and internal gating mechanism of the LSTM cells enable our model to memorize the spatial dependencies and selectively propagate relevant contextual information through the network. We demonstrate improved performance compared to the baseline algorithm with no LSTM units and promising results compared to state-of-the-art methods on Market-1501, CUHK03 and VIPeR datasets. Visualization of the internal mechanism of LSTM cells shows meaningful patterns can be learned by our method.

cs.CV↗

arXiv2016

Spatio-Temporal LSTM with Trust Gates for 3D Human Action Recognition

3D action recognition - analysis of human actions based on 3D skeleton data - becomes popular recently due to its succinctness, robustness, and view-invariant representation. Recent attempts on this problem suggested to develop RNN-based learning methods to model the contextual dependency in the temporal domain. In this paper, we extend this idea to spatio-temporal domains to analyze the hidden sources of action-related information within the input data over both domains concurrently. Inspired by the graphical structure of the human skeleton, we further propose a more powerful tree-structure based traversal method. To handle the noise and occlusion in 3D skeleton data, we introduce new gating mechanism within LSTM to learn the reliability of the sequential input data and accordingly adjust its effect on updating the long-term context information stored in the memory cell. Our method achieves state-of-the-art performance on 4 challenging benchmark datasets for 3D human action analysis.

cs.CV↗

arXiv2016

Full-Time Supervision based Bidirectional RNN for Factoid Question Answering

Recently, bidirectional recurrent neural network (BRNN) has been widely used for question answering (QA) tasks with promising performance. However, most existing BRNN models extract the information of questions and answers by directly using a pooling operation to generate the representation for loss or similarity calculation. Hence, these existing models don't put supervision (loss or similarity calculation) at every time step, which will lose some useful information. In this paper, we propose a novel BRNN model called full-time supervision based BRNN (FTS-BRNN), which can put supervision at every time step. Experiments on the factoid QA task show that our FTS-BRNN can outperform other baselines to achieve the state-of-the-art accuracy.

cs.CL↗

arXiv2016

Dimensionality-Dependent Generalization Bounds for $k$-Dimensional Coding Schemes

The $k$-dimensional coding schemes refer to a collection of methods that attempt to represent data using a set of representative $k$-dimensional vectors, and include non-negative matrix factorization, dictionary learning, sparse coding, $k$-means clustering and vector quantization as special cases. Previous generalization bounds for the reconstruction error of the $k$-dimensional coding schemes are mainly dimensionality independent. A major advantage of these bounds is that they can be used to analyze the generalization error when data is mapped into an infinite- or high-dimensional feature space. However, many applications use finite-dimensional data features. Can we obtain dimensionality-dependent generalization bounds for $k$-dimensional coding schemes that are tighter than dimensionality-independent bounds when data is in a finite-dimensional feature space? The answer is positive. In this paper, we address this problem and derive a dimensionality-dependent generalization bound for $k$-dimensional coding schemes by bounding the covering number of the loss function class induced by the reconstruction error. The bound is of order $\mathcal{O}\left(\left(mk\ln(mkn)/n\right)^{λ_n}\right)$, where $m$ is the dimension of features, $k$ is the number of the columns in the linear implementation of coding schemes, $n$ is the size of sample, $λ_n>0.5$ when $n$ is finite and $λ_n=0.5$ when $n$ is infinite. We show that our bound can be tighter than previous results, because it avoids inducing the worst-case upper bound on $k$ of the loss function and converges faster. The proposed generalization bound is also applied to some specific coding schemes to demonstrate that the dimensionality-dependent bound is an indispensable complement to these dimensionality-independent generalization bounds.

stat.ML↗

arXiv2016

Optical Transients Powered by Magnetars: Dynamics, Light Curves, and Transition to the Nebular Phase

Millisecond magnetars can be formed via several channels: core-collapse of massive stars, accretion-induced collapse of white dwarfs (WDs), double WD mergers, double neutron star (NS) mergers, and WD-NS mergers. Because the mass of ejecta from these channels could be quite different, their light curves are also expected to be diverse. We evaluate the dynamic evolution of optical transients powered by millisecond magnetars. We find that the magnetar with short spin-down timescale converts its rotational energy mostly into the kinetic energy of the transient, while the energy of a magnetar with long spin-down timescale goes into radiation of the transient. This leads us to speculate that hypernovae could be powered by magnetars with short spin-down timescales. At late times the optical transients will gradually evolve into a nebular phase because of the photospheric recession. We treat the photosphere and nebula separately because their radiation mechanisms are different. In some cases the ejecta could be light enough that the magnetar can accelerate it to a relativistic speed. It is well known that the peak luminosity of a supernova (SN) occurs when the luminosity is equal to the instantaneous energy input rate, as shown by Arnett (1979). We show that photospheric recession and relativistic motion can modify this law. The photospheric recession always leads to a delay of the peak time $t_{\mathrm{pk}}$ relative to the time $t_{\times }$ at which the SN luminosity equals the instantaneous energy input rate. Relativistic motion, however, may change this result significantly.

astro-ph.HE↗

arXiv2015

The optical identifcation of events with poorly defined locations: The case of the Fermi GBM GRB140801A

We report the early discovery of the optical afterglow of gamma-ray burst (GRB) 140801A in the 137 deg$^2$ 3-$σ$ error-box of the Fermi Gamma-ray Burst Monitor (GBM). MASTER is the only observatory that automatically react to all Fermi alerts. GRB 140801A is one of the few GRBs whose optical counterpart was discovered solely from its GBM localization. The optical afterglow of GRB 140801A was found by MASTER Global Robotic Net 53 sec after receiving the alert, making it the fastest optical detection of a GRB from a GBM error-box. Spectroscopy obtained with the 10.4-m Gran Telescopio Canarias and the 6-m BTA of SAO RAS reveals a redshift of $z=1.32$. We performed optical and near-infrared photometry of GRB 140801A using different telescopes with apertures ranging from 0.4-m to 10.4-m. GRB 140801A is a typical burst in many ways. The rest-frame bolometric isotropic energy release and peak energy of the burst is $E_\mathrm{iso} = 5.54_{-0.24}^{+0.26} \times 10^{52}$ erg and $E_\mathrm{p, rest}\simeq280$ keV, respectively, which is consistent with the Amati relation. The absence of a jet break in the optical light curve provides a lower limit on the half-opening angle of the jet $θ=6.1$ deg. The observed $E_\mathrm{peak}$ is consistent with the limit derived from the Ghirlanda relation. The joint Fermi GBM and Konus-Wind analysis shows that GRB 140801A could belong to the class of intermediate duration. The rapid detection of the optical counterpart of GRB 140801A is especially important regarding the upcoming experiments with large coordinate error-box areas.

astro-ph.HE↗

arXiv2015

Scalable Nuclear-norm Minimization by Subspace Pursuit Proximal Riemannian Gradient

Nuclear-norm regularization plays a vital role in many learning tasks, such as low-rank matrix recovery (MR), and low-rank representation (LRR). Solving this problem directly can be computationally expensive due to the unknown rank of variables or large-rank singular value decompositions (SVDs). To address this, we propose a proximal Riemannian gradient (PRG) scheme which can efficiently solve trace-norm regularized problems defined on real-algebraic variety $\mMLr$ of real matrices of rank at most $r$. Based on PRG, we further present a simple and novel subspace pursuit (SP) paradigm for general trace-norm regularized problems without the explicit rank constraint $\mMLr$. The proposed paradigm is very scalable by avoiding large-rank SVDs. Empirical studies on several tasks, such as matrix completion and LRR based subspace clustering, demonstrate the superiority of the proposed paradigms over existing methods.

cs.LG↗

arXiv2015

SU(3) and Isospin Breaking Effects on B to PPP Amplitudes

Several modes of $B$ decays into three pseudoscalar octet mesons PPP have been measured. These decays have provided useful information for B decays in the standard model (SM). Some of powerful tools in analyzing B decays are flavor $SU(3)$ and isospin symmetries. Such analyses are usually hampered by $SU(3)$ breaking effects due to a relatively large strange quark mass which breaks SU(3) symmetry down to isospin symmetry. The isospin symmetry also breaks down when up and down quark mass difference is non-zero. It is therefore interesting to find relations which are not sensitive to $SU(3)$ and isospin breaking effects. We find that the relations among several fully-symmetric $B \to PPP$ decay amplitudes are not affected by first order $SU(3)$ breaking effects due to a non-zero strange quark mass, and also some of them are not affected by first isospin breaking effects. These relations, therefore, hold to good precisions. Measurements for these relations can provide important information about B decays in the SM.

hep-ph↗

arXiv2015

PTF11iqb: Cool supergiant mass loss that bridges the gap between Type IIn and normal supernovae

PTF11iqb was initially classified as a TypeIIn event caught very early after explosion. It showed narrow Wolf-Rayet (WR) spectral features on day 2, but the narrow emission weakened quickly and the spectrum morphed to resemble those of Types II-L and II-P. At late times, Halpha emission exhibited a complex, multipeaked profile reminiscent of SN1998S. In terms of spectroscopic evolution, we find that PTF11iqb was a near twin of SN~1998S, although with weaker interaction with circumstellar material (CSM) at early times, and stronger CSM interaction at late times. We interpret the spectral changes as caused by early interaction with asymmetric CSM that is quickly (by day 20) enveloped by the expanding SN ejecta photosphere, but then revealed again after the end of the plateau when the photosphere recedes. The light curve can be matched with a simple model for weak CSM interaction added to the light curve of a normal SN~II-P. This plateau requires that the progenitor had an extended H envelope like a red supergiant, consistent with the slow progenitor wind speed indicated by narrow emission. The cool supergiant progenitor is significant because PTF11iqb showed WR features in its early spectrum --- meaning that the presence of such WR features in an early SN spectrum does not necessarily indicate a WR-like progenitor. [abridged] Overall, PTF11iqb bridges SNe~IIn with weaker pre-SN mass loss seen in SNe II-L and II-P, implying a continuum between these types.

astro-ph.HE↗

arXiv2014

A Continuum of H- to He-Rich Tidal Disruption Candidates With a Preference for E+A Galaxies

We present the results of a Palomar Transient Factory (PTF) archival search for blue transients which lie in the magnitude range between "normal" core-collapse and superluminous supernovae (i.e. with $-21\,{\leq}M_{R\,(peak)}\,{\leq}-19$). Of the six events found after excluding all interacting Type~IIn and Ia-CSM supernovae, three (PTF09ge, 09axc and 09djl) are coincident with the centers of their hosts, one (10iam) is offset from the center, and for two (10nuj and 11glr) a precise offset can not be determined. All the central events have similar rise times to the He-rich tidal disruption candidate PS1-10jh, and the event with the best-sampled light curve also has similar colors and power-law decay. Spectroscopically, PTF09ge is He-rich, while PTF09axc and 09djl display broad hydrogen features around peak magnitude. All three central events are in low star-formation hosts, two of which are E+A galaxies. Our spectrum of the host of PS1-10jh displays similar properties. PTF10iam, the one offset event, is different photometrically and spectroscopically from the central events and its host displays a higher star formation rate. Finding no obvious evidence for ongoing galactic nuclei activity or recent star formation, we conclude that the three central transients likely arise from the tidal disruption of a star by a super-massive black hole. We compare the spectra of these events to tidal disruption candidates from the literature and find that all of these objects can be unified on a continuous scale of spectral properties. The accumulated evidence of this expanded sample strongly supports a tidal disruption origin for this class of nuclear transients.

astro-ph.HE↗

arXiv2014

Exploring the spectral diversity of low-redshift Type Ia supernovae using the Palomar Transient Factory

We present an investigation of the optical spectra of 264 low-redshift (z < 0.2) Type Ia supernovae (SNe Ia) discovered by the Palomar Transient Factory, an untargeted transient survey. We focus on velocity and pseudo-equivalent width measurements of the Si II 4130, 5972, and 6355 A lines, as well those of the Ca II near-infrared (NIR) triplet, up to +5 days relative to the SN B-band maximum light. We find that a high-velocity component of the Ca II NIR triplet is needed to explain the spectrum in ~95 per cent of SNe Ia observed before -5 days, decreasing to ~80 per cent at maximum. The average velocity of the Ca II high-velocity component is ~8500 km/s higher than the photospheric component. We confirm previous results that SNe Ia around maximum light with a larger contribution from the high-velocity component relative to the photospheric component in their Ca II NIR feature have, on average, broader light curves and lower Ca II NIR photospheric velocities. We find that these relations are driven by both a stronger high-velocity component and a weaker contribution from the photospheric Ca II NIR component in broader light curve SNe Ia. We identify the presence of C II in very-early-time SN Ia spectra (before -10 days), finding that >40 per cent of SNe Ia observed at these phases show signs of unburnt material in their spectra, and that C II features are more likely to be found in SNe Ia having narrower light curves.

astro-ph.HE↗