arXiv subjects

Feng Gao

Publications and source records attributed to Feng Gao.

At least 289 records · Page 16Linked to original sources

arXiv2019

RAVEN: A Dataset for Relational and Analogical Visual rEasoNing

Dramatic progress has been witnessed in basic vision tasks involving low-level perception, such as object recognition, detection, and tracking. Unfortunately, there is still an enormous performance gap between artificial vision systems and human intelligence in terms of higher-level vision problems, especially ones involving reasoning. Earlier attempts in equipping machines with high-level reasoning have hovered around Visual Question Answering (VQA), one typical task associating vision and language understanding. In this work, we propose a new dataset, built in the context of Raven's Progressive Matrices (RPM) and aimed at lifting machine intelligence by associating vision with structural, relational, and analogical reasoning in a hierarchical representation. Unlike previous works in measuring abstract reasoning using RPM, we establish a semantic link between vision and reasoning by providing structure representation. This addition enables a new type of abstract reasoning by jointly operating on the structure representation. Machine reasoning ability using modern computer vision is evaluated in this newly proposed dataset. Additionally, we also provide human performance as a reference. Finally, we show consistent improvement across all models by incorporating a simple neural module that combines visual understanding and structure reasoning.

cs.CV↗

arXiv2019

The application of co-integration theory in ensemble pulsar timescale algorithm

Employing multiple pulsars and using an appropriate algorithm to establish ensemble pulsar timescale can reduce the influences of various noises on the long-term stability of pulsar timescale, compared to a single pulsar. However, due to the low timing precision and the significant red noises of some pulsars, their participation in the construction of ensemble pulsar timescale is often limited. Inspired by the principle of solving non-stationary sequence modeling using co-integration theory, we puts forward an algorithm based on the co-integration theory to establish ensemble pulsar timescale. It is found that this algorithm can effectively suppress some noise sources if a co-integration relationship between different pulsar data exist. Different from the classical weighted average algorithm, the co-integration method provides the chances of the pulsar with significant red noises to attend the establishment of ensemble pulsar timescale. Based on the data from the North American Nanohertz Observatory for Gravitational Waves, we found that the co-integration algorithm can successfully reduce several timing noises and improve the long-term stability of the ensemble pulsar timescale.

astro-ph.IM↗

arXiv2019

Geometric Scattering for Graph Data Analysis

We explore the generalization of scattering transforms from traditional (e.g., image or audio) signals to graph data, analogous to the generalization of ConvNets in geometric deep learning, and the utility of extracted graph features in graph data analysis. In particular, we focus on the capacity of these features to retain informative variability and relations in the data (e.g., between individual graphs, or in aggregate), while relating our construction to previous theoretical results that establish the stability of similar transforms to families of graph deformations. We demonstrate the application the our geometric scattering features in graph classification of social network data, and in data exploration of biochemistry data.

cs.LG↗

arXiv2019

Attention-driven Tree-structured Convolutional LSTM for High Dimensional Data Understanding

Modeling the sequential information of image sequences has been a vital step of various vision tasks and convolutional long short-term memory (ConvLSTM) has demonstrated its superb performance in such spatiotemporal problems. Nevertheless, the hierarchical data structures in a significant amount of tasks (e.g., human body parts and vessel/airway tree in biomedical images) cannot be properly modeled by sequential models. Thus, ConvLSTM is not suitable for tree-structured image data analysis. In order to address these limitations, we present tree-structured ConvLSTM models for tree-structured image analysis tasks which can be trained end-to-end. To demonstrate the effectiveness of the proposed tree-structured ConvLSTM model, we present a tree-structured segmentation framework which consists of a tree-structured ConvLSTM and an attention fully convolutional network (FCN) model. The proposed framework is extensively validated on four large-scale coronary artery datasets. The results demonstrate the effectiveness and efficiency of the proposed method.

cs.CV↗

arXiv2018

Residual Attention based Network for Hand Bone Age Assessment

Computerized automatic methods have been employed to boost the productivity as well as objectiveness of hand bone age assessment. These approaches make predictions according to the whole X-ray images, which include other objects that may introduce distractions. Instead, our framework is inspired by the clinical workflow (Tanner-Whitehouse) of hand bone age assessment, which focuses on the key components of the hand. The proposed framework is composed of two components: a Mask R-CNN subnet of pixelwise hand segmentation and a residual attention network for hand bone age assessment. The Mask R-CNN subnet segments the hands from X-ray images to avoid the distractions of other objects (e.g., X-ray tags). The hierarchical attention components of the residual attention subnet force our network to focus on the key components of the X-ray images and generate the final predictions as well as the associated visual supports, which is similar to the assessment procedure of clinicians. We evaluate the performance of the proposed pipeline on the RSNA pediatric bone age dataset and the results demonstrate its superiority over the previous methods.

cs.CV↗

arXiv2018

Angular Softmax Loss for End-to-end Speaker Verification

End-to-end speaker verification systems have received increasing interests. The traditional i-vector approach trains a generative model (basically a factor-analysis model) to extract i-vectors as speaker embeddings. In contrast, the end-to-end approach directly trains a discriminative model (often a neural network) to learn discriminative speaker embeddings; a crucial component is the training criterion. In this paper, we use angular softmax (A-softmax), which is originally proposed for face verification, as the loss function for feature learning in end-to-end speaker verification. By introducing margins between classes into softmax loss, A-softmax can learn more discriminative features than softmax loss and triplet loss, and at the same time, is easy and stable for usage. We make two contributions in this work. 1) We introduce A-softmax loss into end-to-end speaker verification and achieve significant EER reductions. 2) We find that the combination of using A-softmax in training the front-end and using PLDA in the back-end scoring further boosts the performance of end-to-end systems under short utterance condition (short in both enrollment and test). Experiments are conducted on part of $Fisher$ dataset and demonstrate the improvements of using A-softmax.

eess.AS↗

arXiv2018

Research on the Security of Blockchain Data: A Survey

With the more and more extensive application of blockchain, blockchain security has been widely concerned by the society and deeply studied by scholars. Moreover, the security of blockchain data directly affects the security of various applications of blockchain. In this survey, we perform a comprehensive classification and summary of the security of blockchain data. First, we present classification of blockchain data attacks. Subsequently, we present the attacks and defenses of blockchain data in terms of privacy, availability, integrity and controllability. Data privacy attacks present data leakage or data obtained by attackers through analysis. Data availability attacks present abnormal or incorrect access to blockchain data. Data integrity attacks present blockchain data being tampered. Data controllability attacks present blockchain data accidentally manipulated by smart contract vulnerability. Finally, we present several important open research directions to identify follow-up studies in this area.

cs.CR↗

arXiv2018

Multiple Star Systems in the Orion Nebula

This work presents an interferometric study of the massive-binary fraction in the Orion Trapezium Cluster with the recently comissioned GRAVITY instrument. We observe a total of 16 stars of mainly OB spectral type. We find three previously unknown companions for $θ^1$ Ori B, $θ^2$ Ori B, and $θ^2$ Ori C. We determine a separation for the previously suspected companion of NU Ori. We confirm four companions for $θ^1$ Ori A, $θ^1$ Ori C, $θ^1$ Ori D, and $θ^2$ Ori A, all with substantially improved astrometry and photometric mass estimates. We refine the orbit of the eccentric high-mass binary $θ^1$ Ori C and we are able to derive a new orbit for $θ^1$ Ori D. We find a system mass of 21.7 $M_{\odot}$ and a period of $53$ days. Together with other previously detected companions seen in spectroscopy or direct imaging, eleven of the 16 high-mass stars are multiple systems. We obtain a total number of 22 companions with separations up to 600 AU. The companion fraction of the early B and O stars in our sample is about 2, significantly higher than in earlier studies of mostly OB associations. The separation distribution hints towards a bimodality. Such a bimodality has been previously found in A stars, but rarely in OB binaries, which up to this point have been assumed to be mostly compact with a tail of wider companions. We also do not find a substantial population of equal-mass binaries. The observed distribution of mass ratios declines steeply with mass, and like the direct star counts, indicates that our companions follow a standard power law initial mass function. Again, this is in contrast to earlier findings of flat mass ratio distributions in OB associations. We exclude collision as a dominant formation mechanism but find no clear preference for core accretion or competitive accretion.

astro-ph.SR↗

arXiv2018

A Detection of Sgr A* in the far infrared

We report the first detection of the Galactic Centre massive black hole, Sgr~A*, in the far infrared. Our measurements were obtained with PACS on board the \emph{Herschel} satellite at $100~\mathrm{μm}$ and $160~\mathrm{μm}$. While the warm dust in the Galactic Centre is too bright to allow for a direct detection of Sgr~A*, we measure a significant and simultaneous variation of its flux of $ΔF_{ν\widehat{=}160 ~\mathrm{μm}} = (0.27\pm0.06)~\mathrm{Jy}$ and $ΔF_{ν\widehat{=}100 ~\mathrm{μm}}= (0.16\pm0.10)~\mathrm{Jy}$ during one observation. The significance level of the $160 ~\mathrm{μm}$ band variability is $4.5σ$ and the corresponding $100 ~\mathrm{μm}$ band variability is significant at $1.6σ$. We find no example of an equally significant false positive detection. Conservatively assuming a variability of $25\%$ in the FIR, we can provide upper limits to the flux. Comparing the latter with theoretical models we find that 1D RIAF models have difficulties explaining the observed faintness. However, the upper limits are consistent with modern ALMA and VLA observations. Our upper limits provide further evidence for a spectral peak at $\sim 10^{12} ~ \mathrm{Hz}$ and constrain the number density of $γ\sim 100$ electrons in the accretion disk and or outflow.

astro-ph.GA↗

arXiv2018

What stellar orbit is needed to measure the spin of the Galactic center black hole from astrometric data?

Astrometric and spectroscopic monitoring of individual stars orbiting the supermassive black hole in the Galactic Center offer a promising way to detect general relativistic effects. While low-order effects are expected to be detected following the periastron passage of S2 in Spring 2018, detecting higher-order effects due to black hole spin will require the discovery of closer stars. In this paper, we set out to determine the requirements such a star would have to satisfy to allow the detection of black hole spin. We focus on the instrument GRAVITY, which saw first light in 2016 and which is expected to achieve astrometric accuracies $10-100 μ$as. For an observing campaign with duration $T$ years, $N_{obs}$ total observations, astrometric precision $σ_x$ and normalized black hole spin $χ$, we find that $a_{orb}(1-e^2)^{3/4} \lesssim 300 R_S \sqrt{\frac{T}{4 \text{years}}} \left(\frac{N_{obs}}{120}\right)^{0.25} \sqrt{\frac{10 μas}{σ_x}} \sqrt{\fracχ{0.9}}$ is needed. For $χ=0.9$ and a potential observing campaign with $σ_x = 10 μ$as, 30 observations/year and duration 4-10 years, we expect $\sim 0.1$ star with $K<19$ satisfying this constraint based on the current knowledge about the stellar population in the central 1". We also propose a method through which GRAVITY could potentially measure radial velocities with precision $\sim 50$ km/s. If the astrometric precision can be maintained, adding radial velocity information increases the expected number of stars by roughly a factor of two. While we focus on GRAVITY, the results can also be scaled to parameters relevant for future extremely large telescopes.

astro-ph.GA↗

arXiv2017

All-optical transistor based on Rydberg atom-assisted opto-mechanical system

We study the optical response of double optomechanical cavity system assisted by Rydberg atomic ensembles. And atomic ensembles are only coupled with one side cavity by a single cavity mode. It has been realized that a long-range manipulation for optical properties of hybrid system, by controlling the Rydberg atomic ensembles decoupled with the optomechanical cavity. Switching on the coupling between atoms and cavity mode, the original time reversal symmetry of double cavity structure has been broken. Based on the controlled optical non-reciprocity, we put forward the theoretical schemes of all-optical controlled diode, rectifier and transistor.

physics.optics↗

arXiv2017

Incorporating Intra-Class Variance to Fine-Grained Visual Recognition

Fine-grained visual recognition aims to capture discriminative characteristics amongst visually similar categories. The state-of-the-art research work has significantly improved the fine-grained recognition performance by deep metric learning using triplet network. However, the impact of intra-category variance on the performance of recognition and robust feature representation has not been well studied. In this paper, we propose to leverage intra-class variance in metric learning of triplet network to improve the performance of fine-grained recognition. Through partitioning training images within each category into a few groups, we form the triplet samples across different categories as well as different groups, which is called Group Sensitive TRiplet Sampling (GS-TRS). Accordingly, the triplet loss function is strengthened by incorporating intra-class variance with GS-TRS, which may contribute to the optimization objective of triplet network. Extensive experiments over benchmark datasets CompCar and VehicleID show that the proposed GS-TRS has significantly outperformed state-of-the-art approaches in both classification and retrieval tasks.

cs.CV↗

arXiv2017

Improving Object Detection with Region Similarity Learning

Object detection aims to identify instances of semantic objects of a certain class in images or videos. The success of state-of-the-art approaches is attributed to the significant progress of object proposal and convolutional neural networks (CNNs). Most promising detectors involve multi-task learning with an optimization objective of softmax loss and regression loss. The first is for multi-class categorization, while the latter is for improving localization accuracy. However, few of them attempt to further investigate the hardness of distinguishing different sorts of distracting background regions (i.e., negatives) from true object regions (i.e., positives). To improve the performance of classifying positive object regions vs. a variety of negative background regions, we propose to incorporate triplet embedding into learning objective. The triplet units are formed by assigning each negative region to a meaningful object class and establishing class- specific negatives, followed by triplets construction. Over the benchmark PASCAL VOC 2007, the proposed triplet em- bedding has improved the performance of well-known FastRCNN model with a mAP gain of 2.1%. In particular, the state-of-the-art approach OHEM can benefit from the triplet embedding and has achieved a mAP improvement of 1.2%.

cs.CV↗

arXiv2017

A new fractional derivative involving the normalized sinc function without singular kernel

In this paper, a new fractional derivative involving the normalized sinc function without singular kernel is proposed. The Laplace transform is used to find the analytical solution of the anomalous heat-diffusion problems. The comparative results between classical and fractional-order operators are presented. The results are significant in the analysis of one-dimensional anomalous heat-transfer problems.

math.CA↗

arXiv2016

Megamaser Disks Reveal a Broad Distribution of Black Hole Mass in Spiral Galaxies

We use new precision measurements of black hole masses from water megamaser disks to investigate scaling relations between macroscopic galaxy properties and supermassive black hole (BH) mass. The megamaser-derived BH masses span 10^6-10^8 M_sun, while all the galaxy properties that we examine (including stellar mass, central mass density, central velocity dispersion) lie within a narrow range. Thus, no galaxy property correlates tightly with M_BH in ~L* spiral galaxies. Of them all, stellar velocity dispersion provides the tightest relation, but at fixed sigma* the mean megamaser M_BH are offset by -0.6+/-0.1 dex relative to early-type galaxies. Spiral galaxies with non-maser dynamical BH masses do not show this offset. At low mass, we do not yet know the full distribution of BH mass at fixed galaxy property; the non-maser dynamical measurements may miss the low-mass end of the BH distribution due to inability to resolve the spheres of influence and/or megamasers may preferentially occur in lower-mass BHs.

astro-ph.GA↗

arXiv2016

Valley Trion Dynamics in Monolayer MoSe$_2$

Charged excitons called trions play an important role in the fundamental valley dynamics in the newly emerging 2D semiconductor materials. We used ultrafast pump- probe spectroscopy to study the valley trion dynamics in a MoSe$_2$ monolayer grown by using chemical vapor deposition. The dynamics display an ultrafast trion formation followed by a non-exponential decay. The measurements at different pump fluences show that the trion decay dynamics become slower as the excitation density increases. The observed trion dynamics and the associated density dependence are a result of the trapping by two defect states as being the dominating decay mechanism. The simulation based on a set of rate equations reproduces the experimental data for different pump fluences. Our results reveal the important trion dynamics and identify the trapping by defect states as the primary trion decay mechanism in monolayer MoSe$_2$ under the excitation densities used in our experiment.

cond-mat.mes-hall↗

arXiv2015

Imprints of relic gravitational waves on pulsar timing

Relic gravitational waves (RGWs) , a background originated during inflation, would give imprints on the pulsar timing residuals. This makes RGWs be one of important sources for detection using the method of pulsar timing. In this paper, we discuss the effects of RGWs on the single pulsar timing, and give quantitively the timing residuals caused by RGWs with different model parameters. In principle, if the RGWs are strong enough today, they can be detected by timing a single millisecond pulsar with high precision after the intrinsic red noise in pulsar timing residuals were understood, even though observing simultaneously multiple millisecond pulsars is a more powerful technique in extracting gravitational wave signals. We corrected the normalization of RGWs using observations of the cosmic microwave background (CMB), which leads to the amplitudes of RGWs being reduced by two orders of magnitude or so compared to our previous works. We made new constraints on RGWs using the recent observations from the Parkes Pulsar Timing Array, employing the tensor-to-scalar ratio $r=0.2$ due to the tensor-type polarization observations of CMB by BICEP2 as a referenced value even though it has been denied. Moreover, the constraints on RGWs from CMB and BBN (Big Bang nucleosynthesis) will also be discussed for comparison.

gr-qc↗

arXiv2015

Longitudinal Antigenic Sequences and Sites from Intra-Host Evolution (LASSIE) Identifies Immune-Selected HIV Variants

Within-host genetic sequencing from samples collected over time provides a dynamic view of how viruses evade host immunity. Immune-driven mutations might stimulate neutralization breadth by selecting antibodies adapted to cycles of immune escape that generate within-subject epitope diversity. Comprehensive identification of immune-escape mutations is experimentally and computationally challenging. With current technology, many more viral sequences can readily be obtained than can be tested for binding and neutralization, making down-selection necessary. Typically, this is done manually, by picking variants that represent different time-points and branches on a phylogenetic tree. Such strategies are likely to miss many relevant mutations and combinations of mutations, and to be redundant for other mutations. Longitudinal Antigenic Sequences and Sites from Intrahost Evolution (LASSIE) uses transmitted-founder loss to identify virus "hot-spots" under putative immune selection and chooses sequences that represent recurrent mutations in selected sites. LASSIE favors earliest sequences in which mutations arise. With well-characterized longitudinal Env sequences, we confirmed selected sites were concentrated in antibody contacts and selected sequences represented diverse antigenic phenotypes. Practical applications include rapidly identifying immune targets under selective pressure within a subject, selecting minimal sets of reagents for immunological assays that characterize evolving antibody responses, and for immunogens in polyvalent "cocktail" vaccines.

q-bio.PE↗