arXiv subjects

Jun Yang

Publications and source records attributed to Jun Yang.

At least 451 records · Page 25Linked to original sources

arXiv2017

Towards well-specified semi-supervised model-based classifiers via structural adaptation

Semi-supervised learning plays an important role in large-scale machine learning. Properly using additional unlabeled data (largely available nowadays) often can improve the machine learning accuracy. However, if the machine learning model is misspecified for the underlying true data distribution, the model performance could be seriously jeopardized. This issue is known as model misspecification. To address this issue, we focus on generative models and propose a criterion to detect the onset of model misspecification by measuring the performance difference between models obtained using supervised and semi-supervised learning. Then, we propose to automatically modify the generative models during model training to achieve an unbiased generative model. Rigorous experiments were carried out to evaluate the proposed method using two image classification data sets PASCAL VOC'07 and MIR Flickr. Our proposed method has been demonstrated to outperform a number of state-of-the-art semi-supervised learning approaches for the classification task.

cs.LG↗

arXiv2017

Identify Critical Branches with Cascading Failure Chain Statistics and Hypertext-Induced Topic Search Algorithm

An effective way to suppress the cascading failure risk is the branch capacity upgrade, whose optimal decision making, however, may incur high computational burden. A practical way is to find out some critical branches as the candidates in advance. This paper proposes a simulation data oriented approach to identify the critical branches with higher importance in cascading failure propagation. First, a concept of cascading failure chain (CFC) is introduced and numerous samples of CFC are generated with an AC power flow based cascading failure simulator. Then, a directed weighted graph is constructed, whose edges denotes the severities of branch interactions. Third, the weighted hypertext-induced topic search (HITS) algorithm is used to rate and rank this graph's vertices,through which the critical branches can be identified accordingly. Validations on IEEE 118bus and RTS96 systems show that the proposed approach can identify critical branches whose capacity upgrades suppress cascading failure risk more greatly. Moreover, it is also shown that structural importance of a branch does not agree with its importance in cascading failure, which indicates the effectiveness of the proposed approach compared with structure vulnerabilities based identifying methods.

eess.SY↗

arXiv2017

Determination of the Magnetic Fields of Magellanic X-Ray Pulsars

The 80 high-mass X-ray binary (HMXB) pulsars that are known to reside in the Magellanic Clouds (MCs) have been observed by the XMM-Newton and Chandra X-ray telescopes on a regular basis for 15 years, and the XMM-Newton and Chandra archives contain nearly complete information about the duty cycles of the sources with spin periods P_S < 100 s. We have rerprocessed the archival data from both observatories and we combined the output products with all the published observations of 31 MC pulsars with P_S < 100 s in an attempt to investigate the faintest X-ray emission states of these objects that occur when accretion to the polar caps proceeds at the smallest possible rates. These states determine the so-called propeller lines of the accreting pulsars and yield information about the magnitudes of their surface magnetic fields. We have found that the faintest states of the pulsars segregate into five discrete groups which obey to a high degree of accuracy the theoretical relation between spin period and X-ray luminosity. So the entire population of these pulsars can be described by just five propeller lines and the five corresponding magnetic moments (0.29, 0.53, 1.2, 2.9, and 7.3, in units of 10^30 G cm^3).

astro-ph.HE↗

arXiv2017

Design of an impedance matching acoustic bend

We propose the design of an impedance matching acoustic bend in this article. The bending structure is composed of sub-wavelength unit cells with perforated plates and side pipes, whose mass density and bulk modulus can be tuned simultaneously. So the refraction index and the impedance of the acoustic bend can be modulated simultaneously to guarantee both the bending effect and the high transmission. The simulation results of sound pressure field distribution show that the bending effect of the impedance matching acoustic bend is very good. Transmission spectra of the impedance matching acoustic bend and the acoustic bend composed of perforated plates only are both calculated for comparison. The results indicate that the impedance matching acoustic bend is successful in improving the impedance ratio and increasing the transmission obviously.

physics.class-ph↗

arXiv2017

Design and demonstration of an acoustic right-angle bend

In this paper, we design, fabricate and experimentally characterize a broadband acoustic right-angle bend device in air. Perforated panels with various hole-sizes are used to construct the bend structure. Both the simulated and the experimental results verify that acoustic beam can be rotated effectively through the acoustic bend in a wide frequency range. This model may have potential applications in some areas such as sound absorption and acoustic detection in pipeline.

physics.class-ph↗

arXiv2017

A general second order complete active space self-consistent-field solver for large-scale systems

We present a new second order complete active space self-consistent field implementation to converge wavefunctions for both large active spaces and large atomic orbital (AO) bases. Our algorithm decouples the active space wavefunction solver from the orbital optimization in the microiterations, and thus may be easily combined with various modern active space solvers. We also introduce efficient approximate orbital gradient and Hessian updates, and step size determination. We demonstrate its capabilities by calculating the low-lying states of the Fe(\Roman{2})-porphine complex with modest resources using a density matrix renormalization group solver in a CAS(22,27) active space and a 3000 AO basis.

physics.chem-ph↗

arXiv2017

Broadband solid cloak for underwater acoustics

Application of transformation theory to underwater acoustics has been a challenging task because highly anisotropic density is unachievable in water. A possible strategy is to exploit anisotropic modulus rather than density, while has not been experimentally demonstrated. We present an annular underwater acoustic cloak designed from particular graded solid microstructures. The geometry tailored microstructures mimics meta-fluid with highly anisotropic modulus through substantially suppressed shear wave. Transient wave experiments are conducted with the cloak in a designed 2D underwater waveguide system and proved excellent cloaking performance for enclosed target over broadband frequency 9-15 kHz. This finding paves the way for controlling underwater acoustics using the structured anisotropic modulus meta-fluid.

physics.class-ph↗

arXiv2016

Discovery of five low luminosity active galactic nuclei at the centre of the Perseus cluster

According to optical stellar kinematics observations, an over-massive black hole candidate has been reported by van den Bosch et al. (2012) in the normal early-type galaxy NGC 1277. This galaxy is located in the central region of the Perseus cluster. Westerbork Synthesis Radio Telescope (WSRT) observations have shown that NGC 1277 and other early-type galaxies in the neighbourhood have radio counterparts. These nuclear radio sources have stable flux densities on time scale of years. In order to investigate the origin of the radio emission from these normal galaxies, we selected five sources (NGC 1270, NGC 1272, NGC 1277, NGC 1278 and VZw 339) residing in the central 10 arcminute region of the Perseus cluster and requested to re-correlate the data of an existing very long baseline interferometry (VLBI) experiment at these new positions. With the re-correlation data provided by the European VLBI Network (EVN), we imaged the five sources with a resolution of about eight milliarcseconds and detected all of them with a confidence level above 5σ at 1.4 GHz. They show compact structure and brightness temperatures above $10^7$ K, which implies that the radio emission is non-thermal. We rule out ongoing nuclear star formation and conclude that these VLBI-detected radio sources are parsec-scale jet activity associated with the supermassive black holes in low luminosity active galactic nuclei (LLAGNs), although there are no clear signs of nuclear activity observed in the optical and infrared bands. Using the fundamental plane relation in black holes, we find no significant evidence for or against an extremely massive black hole hiding in NGC 1277.

astro-ph.GA↗

arXiv2016

Unsupervised Co-segmentation of 3D Shapes via Functional Maps

We present an unsupervised method for co-segmentation of a set of 3D shapes from the same class with the aim of segmenting the input shapes into consistent semantic parts and establishing their correspondence across the set. Starting from meaningful pre-segmentation of all given shapes individually, we construct the correspondence between same candidate parts and obtain the labels via functional maps. And then, we use these labels to mark the input shapes and obtain results of co-segmentation. The core of our algorithm is to seek for an optimal correspondence between semantically similar parts through functional maps and mark such shape parts. Experimental results on the benchmark datasets show the efficiency of this method and comparable accuracy to the state-of-the-art algorithms.

cs.GR↗

arXiv2016

NGC5252: a pair of radio-emitting active galactic nuclei?

The X-ray source CXO J133815.6+043255 has counterparts in the UV, optical, and radio bands. Based on the multi-band investigations, it has been recently proposed by Kim et al. (2015) as a rarely-seen off-nucleus ultraluminous X-ray (ULX) source with a black hole mass of >= 10^4 solar mass in the nearby Seyfert galaxy NGC 5252. To explore its radio properties at very high angular resolution, we performed very long-baseline interferometry (VLBI) observations with the European VLBI Network (EVN) at 1.7 GHz. We find that the radio counterpart is remarkably compact among the known ULXs. It does not show a resolved structure with a resolution of a few milliarcsecond (mas), and the total recovered flux density is comparable to that measured in earlier sub-arcsecond-resolution images. The compact radio structure, the relatively flat spectrum, and the high radio luminosity are consistent with a weakly accreting supermassive black hole in a low-luminosity active galactic nucleus. The nucleus of NGC 5252 itself has similar radio properties. We argue that the system represents a relatively rare pair of active galactic nuclei, where both components emit in the radio.

astro-ph.HE↗

arXiv2016

The most-luminous heavily-obscured quasars have a high merger fraction: morphological study of WISE-selected hot dust-obscured galaxies

Previous studies have shown that WISE-selected hyperluminous, hot dust-obscured galaxies (Hot DOGs) are powered by highly dust-obscured, possibly Compton-thick AGNs. High obscuration provides us a good chance to study the host morphology of the most luminous AGNs directly. We analyze the host morphology of 18 Hot DOGs at $z\sim3$ using Hubble Space Telescope/WFC3 imaging. We find that Hot DOGs have a high merger fraction ($62\pm 14 \%$). By fitting the surface brightness profiles, we find that the distribution of Sérsic indices in our Hot DOG sample peaks around 2, which suggests that most of Hot DOGs have transforming morphologies. We also derive the AGN bolometric luminosity ($\sim10^{14}L_\odot$) of our Hot DOG sample by using IR SEDs decomposition. The derived merger fraction and AGN bolometric luminosity relation is well consistent with the variability-based model prediction (Hickox et al. 2014). Both the high merger fraction in IR-luminous AGN sample and relatively low merger fraction in UV/optical-selected, unobscured AGN sample can be expected in the merger-driven evolutionary model. Finally, we conclude that Hot DOGs are merger-driven and may represent a transit phase during the evolution of massive galaxies, transforming from the dusty starburst dominated phase to the unobscured QSO phase.

astro-ph.GA↗

arXiv2016

On Ambrosetti-Malchiodi-Ni conjecture on two-dimensional smooth bounded domains

We consider the problem $$ ε^2 Δu-V(y)u+u^p\,=\,0,~~u>0~~\quad\mbox{in}\quadΩ,~~\quad\frac {\partial u}{\partial ν}\,=\,0\quad\mbox{on}~~~\partial Ω, $$ where $Ω$ is a bounded domain in $\mathbb R^2$ with smooth boundary, the exponent $p>1$, $ε>0$ is a small parameter, $V$ is a uniformly positive, smooth potential on $\barΩ$, and $ν$ denotes the outward normal of $\partial Ω$. Let $Γ$ be a curve intersecting orthogonally with $\partial Ω$ at exactly two points and dividing $Ω$ into two parts. Moreover, $Γ$ satisfies stationary and non-degeneracy conditions with respect to the functional $\int_ΓV^σ$, where $σ=\frac {p+1}{p-1}-\frac 12$. We prove the existence of a solution $u_ε$ concentrating along the whole of $Γ$, exponentially small in $ε$ at any positive distance from it, provided that $ε$ is small and away from certain critical numbers. In particular, this establishes the validity of the two dimensional case of a conjecture by A. Ambrosetti, A. Malchiodi and W.-M. Ni(p.327, [4]).

math.AP↗

arXiv2016

The inner edge of the habitable zone for synchronously rotating planets around low-mass stars using general circulation models

Terrestrial planets at the inner edge of the habitable zone of late-K and M-dwarf stars are expected to be in synchronous rotation, as a consequence of strong tidal interactions with their host stars. Previous global climate model (GCM) studies have shown that, for slowly-rotating planets, strong convection at the substellar point can create optically thick water clouds, increasing the planetary albedo, and thus stabilizing the climate against a thermal runaway. However these studies did not use self-consistent orbital/rotational periods for synchronously rotating planets placed at different distances from the host star. Here we provide new estimates of the inner edge of the habitable zone for synchronously rotating terrestrial planets around late-K and M-dwarf stars using a 3-D Earth-analog GCM with self-consistent relationships between stellar metallicity, stellar effective temperature, and the planetary orbital/rotational period. We find that both atmospheric dynamics and the efficacy of the substellar cloud deck are sensitive to the precise rotation rate of the planet. Around mid-to-late M-dwarf stars with low metallicity, planetary rotation rates at the inner edge of the HZ become faster, and the inner edge of the habitable zone is farther away from the host stars than in previous GCM studies. For an Earth-sized planet, the dynamical regime of the substellar clouds begins to transition as the rotation rate approaches ~10 days. These faster rotation rates produce stronger zonal winds that encircle the planet and smear the substellar clouds around it, lowering the planetary albedo, and causing the onset of the water-vapor greenhouse climatic instability to occur at up to ~25% lower incident stellar fluxes than found in previous GCM studies. For mid-to-late M-dwarf stars with high metallicity and for mid-K to early-M stars, we agree with previous studies.

astro-ph.EP↗

arXiv2016

On Bounding the Union Probability Using Partial Weighted Information

Effective bounds on the union probability are well known to be beneficial in the analysis of stochastic problems in many areas, including probability theory, information theory, statistical communications, computing and operations research. In this work we present new results on bounding the probability of a finite union of events, ${P\left(\bigcup_{i=1}^N A_i\right)}$, for a fixed positive integer ${N}$, using partial information on the events in terms of ${\{P(A_i)\}}$ and ${\{\sum_j c_j P(A_i\cap A_j)\}}$ where ${c_1}$, ${\dots}$, ${c_N}$ are given weights. We derive two new classes of lower bounds of at most pseudo-polynomial computational complexity. These classes of lower bounds generalize the existing bound in \cite{Kuai2000} and recent bounds in \cite{Yang2014,Yang2014ISIT} and are numerically shown to be tighter in some cases than the Gallot-Kounias bound \cite{Gallot1966,Kounias1968} and the Pr{é}kopa-Gao bound \cite{Prekopa2005} which require more information on the events probabilities.

math.PR↗

arXiv2015

An acoustic invisible gateway

The recently-emerged concept of "invisible gateway" with the extraordinary capability to block the waves but allow the passage of other entities has attracted great attentions due to the general interests in illusion devices. However, the possibility to realize such a fascinating phenomenon for acoustic waves has not yet been explored, which should be of paramount significance for acoustical applications but would necessarily involve experimental difficulty. Here we design and experimentally demonstrate an acoustic invisible gateway (AIG) capable of concealing a channel under the detection of sound. Instead of "restoring" a whole block of background medium by using transformation acoustics that inevitably requires complementary or restoring media with extreme parameters, we propose an inherently distinct methodology that only aims at engineering the surface impedance at the "gate" to mimic a rigid "wall" and can be conveniently implemented by decorating meta-structures behind the channel. Such a simple yet effective design of AIG also enables analyzing the physics behind this extraordinary phenomenon in an analytical manner, which agrees quite well with the numerical and experimental results. Furthermore, our scheme offers possibility to manipulate the symmetry of AIGs freely, which is proven both theoretically and experimentally by demonstrating two distinctive examples of one-way and two-way AIGs. We anticipate the realization of AIGs will open new avenues to illusion acoustics with potential applications in acoustic measurement and acoustic device, and provide inspirations for similar researches in other fields.

physics.class-ph↗

arXiv2015

VLBI detection of internal shocks in nova V959 Mon

V959 Mon is a classical nova detected at GeV gamma-ray wavelengths on 2012 June 19. While classical novae are now routinely detected in gamma-rays, the origin of the shocks that produce relativistic particles has remained unknown. We carried out electronic European VLBI Network (e-EVN) observations that revealed a pair of compact synchrotron emission features in V959 Mon on 2012 Sep 18. Since synchrotron emission requires strong shocks as well, we identify these features as the location where the gamma rays were produced. We also detected the extended ejecta in the follow-up EVN observations. They expanded much faster in East-West direction than the compact knots detected in the aforementioned e-EVN measurements. By comparing the VLBI results with lower resolution images obtained using e-MERLIN and the VLA - as reported by Chomiuk et al. (2014) - it appears that 1) influenced by the binary orbit, the nova ejecta was highly asymmetric with a dense and slow outflow in the equatorial plane and low-density and faster ejecta along the poles; and 2) the VLBI knots were related to shocks formed in the interaction region of these outflows.

astro-ph.HE↗

arXiv2015

High-quality Image Restoration from Partial Mixed Adaptive-Random Measurements

A novel framework to construct an efficient sensing (measurement) matrix, called mixed adaptive-random (MAR) matrix, is introduced for directly acquiring a compressed image representation. The mixed sampling (sensing) procedure hybridizes adaptive edge measurements extracted from a low-resolution image with uniform random measurements predefined for the high-resolution image to be recovered. The mixed sensing matrix seamlessly captures important information of an image, and meanwhile approximately satisfies the restricted isometry property. To recover the high-resolution image from MAR measurements, the total variation algorithm based on the compressive sensing theory is employed for solving the Lagrangian regularization problem. Both peak signal-to-noise ratio and structural similarity results demonstrate the MAR sensing framework shows much better recovery performance than the completely random sensing one. The work is particularly helpful for high-performance and lost-cost data acquisition.

cs.IT↗

arXiv2014

Very Long Baseline Interferometry with the SKA

Adding VLBI capability to the SKA arrays will greatly broaden the science of the SKA, and is feasible within the current specifications. SKA-VLBI can be initially implemented by providing phased-array outputs for SKA1-MID and SKA1-SUR and using these extremely sensitive stations with other radio telescopes, and in SKA2 by realising a distributed configuration providing baselines up to thousands of km, merging it with existing VLBI networks. The motivation for and the possible realization of SKA-VLBI is described in this paper.

astro-ph.IM↗