arXiv subjects

Yu Rong

Publications and source records attributed to Yu Rong.

At least 181 records · Page 10Linked to original sources

arXiv2020

Rumor Detection on Social Media with Bi-Directional Graph Convolutional Networks

Social media has been developing rapidly in public due to its nature of spreading new information, which leads to rumors being circulated. Meanwhile, detecting rumors from such massive information in social media is becoming an arduous challenge. Therefore, some deep learning methods are applied to discover rumors through the way they spread, such as Recursive Neural Network (RvNN) and so on. However, these deep learning methods only take into account the patterns of deep propagation but ignore the structures of wide dispersion in rumor detection. Actually, propagation and dispersion are two crucial characteristics of rumors. In this paper, we propose a novel bi-directional graph model, named Bi-Directional Graph Convolutional Networks (Bi-GCN), to explore both characteristics by operating on both top-down and bottom-up propagation of rumors. It leverages a GCN with a top-down directed graph of rumor spreading to learn the patterns of rumor propagation, and a GCN with an opposite directed graph of rumor diffusion to capture the structures of rumor dispersion. Moreover, the information from the source post is involved in each layer of GCN to enhance the influences from the roots of rumors. Encouraging empirical results on several benchmarks confirm the superiority of the proposed method over the state-of-the-art approaches.

cs.SI↗

arXiv2019

A Restricted Black-box Adversarial Framework Towards Attacking Graph Embedding Models

With the great success of graph embedding model on both academic and industry area, the robustness of graph embedding against adversarial attack inevitably becomes a central problem in graph learning domain. Regardless of the fruitful progress, most of the current works perform the attack in a white-box fashion: they need to access the model predictions and labels to construct their adversarial loss. However, the inaccessibility of model predictions in real systems makes the white-box attack impractical to real graph learning system. This paper promotes current frameworks in a more general and flexible sense -- we demand to attack various kinds of graph embedding model with black-box driven. To this end, we begin by investigating the theoretical connections between graph signal processing and graph embedding models in a principled way and formulate the graph embedding model as a general graph signal process with corresponding graph filter. As such, a generalized adversarial attacker: GF-Attack is constructed by the graph filter and feature matrix. Instead of accessing any knowledge of the target classifiers used in graph embedding, GF-Attack performs the attack only on the graph filter in a black-box attack fashion. To validate the generalization of GF-Attack, we construct the attacker on four popular graph embedding models. Extensive experimental results validate the effectiveness of our attacker on several benchmark datasets. Particularly by using our attack, even small graph perturbations like one-edge flip is able to consistently make a strong attack in performance to different graph embedding models.

cs.SI↗

arXiv2019

Graph Convolutional Networks for Temporal Action Localization

Most state-of-the-art action localization systems process each action proposal individually, without explicitly exploiting their relations during learning. However, the relations between proposals actually play an important role in action localization, since a meaningful action always consists of multiple proposals in a video. In this paper, we propose to exploit the proposal-proposal relations using Graph Convolutional Networks (GCNs). First, we construct an action proposal graph, where each proposal is represented as a node and their relations between two proposals as an edge. Here, we use two types of relations, one for capturing the context information for each proposal and the other one for characterizing the correlations between distinct actions. Then we apply the GCNs over the graph to model the relations among different proposals and learn powerful representations for the action classification and localization. Experimental results show that our approach significantly outperforms the state-of-the-art on THUMOS14 (49.1% versus 42.8%). Moreover, augmentation experiments on ActivityNet also verify the efficacy of modeling action proposal relationships. Codes are available at https://github.com/Alvin-Zeng/PGCN.

cs.CV↗

arXiv2019

Delving Deep Into Hybrid Annotations for 3D Human Recovery in the Wild

Though much progress has been achieved in single-image 3D human recovery, estimating 3D model for in-the-wild images remains a formidable challenge. The reason lies in the fact that obtaining high-quality 3D annotations for in-the-wild images is an extremely hard task that consumes enormous amount of resources and manpower. To tackle this problem, previous methods adopt a hybrid training strategy that exploits multiple heterogeneous types of annotations including 3D and 2D while leaving the efficacy of each annotation not thoroughly investigated. In this work, we aim to perform a comprehensive study on cost and effectiveness trade-off between different annotations. Specifically, we focus on the challenging task of in-the-wild 3D human recovery from single images when paired 3D annotations are not fully available. Through extensive experiments, we obtain several observations: 1) 3D annotations are efficient, whereas traditional 2D annotations such as 2D keypoints and body part segmentation are less competent in guiding 3D human recovery. 2) Dense Correspondence such as DensePose is effective. When there are no paired in-the-wild 3D annotations available, the model exploiting dense correspondence can achieve 92% of the performance compared to a model trained with paired 3D data. We show that incorporating dense correspondence into in-the-wild 3D human recovery is promising and competitive due to its high efficiency and relatively low annotating cost. Our model trained with dense correspondence can serve as a strong reference for future research.

cs.CV↗

arXiv2019

The Next Generation Fornax Survey (NGFS): VI. The Alignment of Dwarf Galaxies in The Fornax Cluster

Using the photometric data from the Next Generation Fornax Survey, we find a significant radial alignment signal among the Fornax dwarf galaxies. For the first time, we report that the radial alignment signal of nucleated dwarfs is stronger than that of non-nucleated ones at 2.4$σ$ confidence level, and the dwarfs located in the outer region ($R>R_{\rm{vir}}/3$; $R_{\rm{vir}}$ is the Fornax virial radius) show slightly stronger radial alignment signal than those in the inner region ($R<R_{\rm{vir}}/3$) at $1.5σ$ level. We also find that the significance of radial alignment signal is independent of the luminosities or sizes of the dwarfs.

astro-ph.GA↗

arXiv2019

Unsupervised Adversarial Graph Alignment with Graph Embedding

Graph alignment, also known as network alignment, is a fundamental task in social network analysis. Many recent works have relied on partially labeled cross-graph node correspondences, i.e., anchor links. However, due to the privacy and security issue, the manual labeling of anchor links for diverse scenarios may be prohibitive. Aligning two graphs without any anchor links is a crucial and challenging task. In this paper, we propose an Unsupervised Adversarial Graph Alignment (UAGA) framework to learn a cross-graph alignment between two embedding spaces of different graphs in a fully unsupervised fashion (\emph{i.e.,} no existing anchor links and no users' personal profile or attribute information is available). The proposed framework learns the embedding spaces of each graph, and then attempts to align the two spaces via adversarial training, followed by a refinement procedure. We further extend our UAGA method to incremental UAGA (iUAGA) that iteratively reveals the unobserved user links based on the pseudo anchor links. This can be used to further improve both the embedding quality and the alignment accuracy. Moreover, the proposed methods will benefit some real-world applications, \emph{e.g.,} link prediction in social networks. Comprehensive experiments on real-world data demonstrate the effectiveness of our proposed approaches UAGA and iUAGA for unsupervised graph alignment.

cs.SI↗

arXiv2019

Progressive Feature Alignment for Unsupervised Domain Adaptation

Unsupervised domain adaptation (UDA) transfers knowledge from a label-rich source domain to a fully-unlabeled target domain. To tackle this task, recent approaches resort to discriminative domain transfer in virtue of pseudo-labels to enforce the class-level distribution alignment across the source and target domains. These methods, however, are vulnerable to the error accumulation and thus incapable of preserving cross-domain category consistency, as the pseudo-labeling accuracy is not guaranteed explicitly. In this paper, we propose the Progressive Feature Alignment Network (PFAN) to align the discriminative features across domains progressively and effectively, via exploiting the intra-class variation in the target domain. To be specific, we first develop an Easy-to-Hard Transfer Strategy (EHTS) and an Adaptive Prototype Alignment (APA) step to train our model iteratively and alternatively. Moreover, upon observing that a good domain adaptation usually requires a non-saturated source classifier, we consider a simple yet efficient way to retard the convergence speed of the source classification loss by further involving a temperature variate into the soft-max function. The extensive experimental results reveal that the proposed PFAN exceeds the state-of-the-art performance on three UDA datasets.

cs.CV↗

arXiv2019

Semi-Supervised Graph Classification: A Hierarchical Graph Perspective

Node classification and graph classification are two graph learning problems that predict the class label of a node and the class label of a graph respectively. A node of a graph usually represents a real-world entity, e.g., a user in a social network, or a protein in a protein-protein interaction network. In this work, we consider a more challenging but practically useful setting, in which a node itself is a graph instance. This leads to a hierarchical graph perspective which arises in many domains such as social network, biological network and document collection. For example, in a social network, a group of people with shared interests forms a user group, whereas a number of user groups are interconnected via interactions or common members. We study the node classification problem in the hierarchical graph where a `node' is a graph instance, e.g., a user group in the above example. As labels are usually limited in real-world data, we design two novel semi-supervised solutions named \underline{SE}mi-supervised gr\underline{A}ph c\underline{L}assification via \underline{C}autious/\underline{A}ctive \underline{I}teration (or SEAL-C/AI in short). SEAL-C/AI adopt an iterative framework that takes turns to build or update two classifiers, one working at the graph instance level and the other at the hierarchical graph level. To simplify the representation of the hierarchical graph, we propose a novel supervised, self-attentive graph embedding method called SAGE, which embeds graph instances of arbitrary size into fixed-length vectors. Through experiments on synthetic data and Tencent QQ group data, we demonstrate that SEAL-C/AI not only outperform competing methods by a significant margin in terms of accuracy/Macro-F1, but also generate meaningful interpretations of the learned representations.

cs.CV↗

arXiv2019

The Next Generation Fornax Survey (NGFS): V. Discovery of a dwarf-dwarf galaxy pair at z=0.30 and its characterization using deep VLT/MUSE observations

We report the detection of a pair of dwarf galaxies at $z\!=\!0.30$ which may be in the early stages of an interaction. Both galaxies have stellar masses of $<10^{9}M_\odot$, and display a projected separation of $\sim\!29$kpc and a physical separation of $\sim\!240$kpc. Evidence of ongoing star formation has been found in both galaxies, with neither one showing an enhanced star-formation rate that would be expected if they were already interacting. One galaxy displays a disturbed morphology but shows ordered gas rotation, which may reflect a previous minor merger event in the recent history of that system. The nearest spectroscopically confirmed neighbour lies at a distance of 38Mpc. These results indicate that these dwarf galaxies have no neighbouring massive galaxies, however with the data available we have been unable to determine whether these galaxies are isolated in the field or belong to a group of low-mass galaxies. As a detection of a rare dwarf-dwarf pair beyond the Local Universe, this system provides an uncommon opportunity to explore the properties of galaxy groups in the low-galaxy mass regime as a function of redshift.

astro-ph.GA↗

arXiv2019

SDSS-IV MaNGA: Inside-out vs. outside-in quenching in different local environments

The large Integral Field Spectroscopy (IFS) surveys have allowed the classification of ionizing sources of emission lines on sub-kpc scales. In this work, we define two non-parametric parameters, quiescence (f$_{q}$) and its concentration (c$_{q}$), to quantify the strength and the spatial distribution of the quenched areas, respectively, traced by the LI(N)ER regions with low EW(H$α$). With these two measurements, we classify MaNGA galaxies into inside-out and outside-in quenching types according to their locations on the f$_{q}$ vs. c$_{q}$ plane and we measure the fraction of inside-out (outside-in) quenching galaxies as a function of halo mass. We find that the fraction of galaxies showing inside-out quenching increases with halo mass, irrespective of stellar mass or galaxy type (satellites vs. centrals). In addition, high stellar mass galaxies exhibit a greater fraction of inside-out quenching compared to low stellar mass ones in all environments. In contrast, the fraction of outside-in quenching does not depend on halo mass. Our results suggest that morphological quenching may be responsible for the inside-out quenching seen in all environments. On the other hand, the flat dependence of the outside-in quenching on halo mass could be a mixed result of ram-pressure stripping and galaxy mergers. Nevertheless, at a given environment and stellar mass, the fraction of inside-out quenching is systematically greater than that of outside-in quenching, suggesting that inside-out quenching is the dominant quenching mode in all environments.

astro-ph.GA↗

arXiv2018

Adaptive Sampling Towards Fast Graph Representation Learning

Graph Convolutional Networks (GCNs) have become a crucial tool on learning representations of graph vertices. The main challenge of adapting GCNs on large-scale graphs is the scalability issue that it incurs heavy cost both in computation and memory due to the uncontrollable neighborhood expansion across layers. In this paper, we accelerate the training of GCNs through developing an adaptive layer-wise sampling method. By constructing the network layer by layer in a top-down passway, we sample the lower layer conditioned on the top one, where the sampled neighborhoods are shared by different parent nodes and the over expansion is avoided owing to the fixed-size sampling. More importantly, the proposed sampler is adaptive and applicable for explicit variance reduction, which in turn enhances the training of our method. Furthermore, we propose a novel and economical approach to promote the message passing over distant nodes by applying skip connections. Intensive experiments on several benchmarks verify the effectiveness of our method regarding the classification accuracy while enjoying faster convergence speed.

cs.CV↗

arXiv2018

The Dawn Light of Blueberry Galaxies: Spectroscopic and Photometric Studies of two Starburst Dwarf Galaxies

A population of so-called "blueberry" starbursting dwarf galaxies with extremely blue colors, low-metallicities, and enormous ionization ratios, has recently been found by Yang et al. (2017). Yet we still do not know their detailed properties, such as morphologies, AGN occupations, massive star contents, infrared emission, dust properties, etc. As a pilot study of the blueberries, we investigate the spectroscopic and photometric properties of two blueberry candidates, RGG B and RGG 5, for which Hubble Space Telescope high-resolution images are available. We find that RGG B and RGG 5 perhaps are likely to be two merging dwarf galaxy systems. RGG B may have a close merging companion; yet the current evidence still cannot exclude the possibility that RGG B is just disturbed by in-situ star formation through, e.g., outflows, rather than undergoing a merger. RGG 5 presents stellar shells in the outskirt which can be the powerful evidence of galaxy merging. We also find that, all of the blueberries, including RGG B and RGG 5, are located close to the theoretical maximum-starburst-line in the BPT diagram, have very high ionization parameters, and relatively low hardness ionizing radiation fields, exhibit nitrogen overabundances, and show extremely red mid-IR colors, and reside in the so-called "ULIRGs/LINERs/Obscured AGN" region. The blueberry galaxies may not harbor AGN.

astro-ph.GA↗

arXiv2018

On the Acceleration of L-BFGS with Second-Order Information and Stochastic Batches

This paper proposes a framework of L-BFGS based on the (approximate) second-order information with stochastic batches, as a novel approach to the finite-sum minimization problems. Different from the classical L-BFGS where stochastic batches lead to instability, we use a smooth estimate for the evaluations of the gradient differences while achieving acceleration by well-scaling the initial Hessians. We provide theoretical analyses for both convex and nonconvex cases. In addition, we demonstrate that within the popular applications of least-square and cross-entropy losses, the algorithm admits a simple implementation in the distributed environment. Numerical experiments support the efficiency of our algorithms.

cs.LG↗

arXiv2018

Pose-Robust Face Recognition via Deep Residual Equivariant Mapping

Face recognition achieves exceptional success thanks to the emergence of deep learning. However, many contemporary face recognition models still perform relatively poor in processing profile faces compared to frontal faces. A key reason is that the number of frontal and profile training faces are highly imbalanced - there are extensively more frontal training samples compared to profile ones. In addition, it is intrinsically hard to learn a deep representation that is geometrically invariant to large pose variations. In this study, we hypothesize that there is an inherent mapping between frontal and profile faces, and consequently, their discrepancy in the deep representation space can be bridged by an equivariant mapping. To exploit this mapping, we formulate a novel Deep Residual EquivAriant Mapping (DREAM) block, which is capable of adaptively adding residuals to the input deep representation to transform a profile face representation to a canonical pose that simplifies recognition. The DREAM block consistently enhances the performance of profile face recognition for many strong deep networks, including ResNet models, without deliberately augmenting training data of profile faces. The block is easy to use, light-weight, and can be implemented with a negligible computational overhead.

cs.CV↗

arXiv2018

SDSS-IV MaNGA: A Distinct Mass Distribution Explored in Slow-Rotating Early-type Galaxies

We study the radial acceleration relation (RAR) for early-type galaxies (ETGs) in the SDSS MaNGA MPL5 dataset. The complete ETG sample show a slightly offset RAR from the relation reported by McGaugh et al. (2016) at the low-acceleration end; we find that the deviation is due to the fact that the slow rotators show a systematically higher acceleration relation than the McGaugh's RAR, while the fast rotators show a consistent acceleration relation to McGaugh's RAR. There is a 1σsignificant difference between the acceleration relations of the fast and slow rotators, suggesting that the acceleration relation correlates with the galactic spins, and that the slow rotators may have a different mass distribution compared with fast rotators and late-type galaxies. We suspect that the acceleration relation deviation of slow rotators may be attributed to more galaxy merger events, which would disrupt the original spins and correlated distributions of baryons and dark matter orbits in galaxies.

astro-ph.GA↗

arXiv2018

SDSS-IV MaNGA: Modelling the metallicity gradients of gas and stars - radially dependent metal outflow vs IMF

In our previous work, we found that only two scenarios are capable of reproducing the observed integrated mass-metallicity relations for the gas and stellar components of local star-forming galaxies simultaneously. One scenario invokes a time-dependent metal outflow loading factor with stronger outflows at early times. The other scenario uses a time-dependent IMF slope with a steeper IMF at early times. In this work, we extend our study to investigate the radial profile of gas and stellar metallicity in local star-forming galaxies using spatially resolved spectroscopic data from the SDSS-IV MaNGA survey. We find that most galaxies show negative gradients in both gas and stellar metallicity with steeper gradients in stellar metallicity. The stellar metallicity gradient tend to be mass dependent with steeper gradients in more massive galaxies while no clear mass dependence is found for the gas metallicity gradient. Then we compare the observations with the predictions from a chemical evolution model of the radial profiles of gas and stellar metallicities. We confirm that the two scenarios proposed in our previous work are also required to explain the metallicity gradients. Based on these two scenarios we successfully reproduce the radial profiles of gas metallicity, stellar metallicity, stellar mass surface density, and star formation rate (SFR) surface density simultaneously. The origin of the negative gradient in stellar metallicity turns out to be driven by either radially dependent metal outflow or IMF slope. In contrast, the radial dependence of the gas metallicity is less constrained because of the degeneracy in model parameters.

astro-ph.GA↗

arXiv2018

The Next Generation Fornax Survey (NGFS): II. The Central Dwarf Galaxy Population

We present a photometric study of the dwarf galaxy population in the core region ($< r_{\rm vir}/4$) of the Fornax galaxy cluster based on deep $u'g'i'$ photometry from the Next Generation Fornax Cluster Survey. All imaging data were obtained with the Dark Energy Camera mounted on the 4-meter Blanco telescope at the Cerro-Tololo Interamerican Observatory. We identify 258 dwarf galaxy candidates with luminosities $-17 < M_{g'} < -8$ mag, corresponding to typical stellar masses of $9.5\gtrsim \log{\cal M}_{\star}/M_\odot \gtrsim 5.5$, reaching $\sim\!3$ mag deeper in point-source luminosity and $\sim\!4$ mag deeper in surface-brightness sensitivity compared to the classic Fornax Cluster Catalog. Morphological analysis shows that surface-brightness profiles are well represented by single-component Sérsic models with average Sérsic indices of $\langle n\rangle_{u',g',i'}=(0.78-0.83) \pm 0.02$, and average effective radii of $\langle r_e\rangle_{u',g',i'}\!=(0.67-0.70) \pm 0.02$ kpc. Color-magnitude relations indicate a flattening of the galaxy red sequence at faint galaxy luminosities, similar to the one recently discovered in the Virgo cluster. A comparison with population synthesis models and the galaxy mass-metallicity relation reveals that the average faint dwarf galaxy is likely older than ~5 Gyr. We study galaxy scaling relations between stellar mass, effective radius, and stellar mass surface density over a stellar mass range covering six orders of magnitude. We find that over the sampled stellar mass range several distinct mechanisms of galaxy mass assembly can be identified: i) dwarf galaxies assemble mass inside the half-mass radius up to $\log{\cal M}_{\star}$ ~8.0, ii) isometric mass assembly in the range $8.0 < \log{\cal M}_{\star}/M_\odot < 10.5$, and iii) massive galaxies assemble stellar mass predominantly in their halos at $\log{\cal M}_{\star}$ ~10.5 and above.

astro-ph.GA↗

arXiv2017

Massive Quiescent Galaxies at z>3 in The Millennium Simulation Populated by A Semi-analytic Galaxy Formation Model

We take advantage of the statistical power of the large-volume dark-matter-only Millennium simulation, combined with a sophisticated semi-analytic galaxy formation model, to explore whether the recently reported $z=3.7$ quiescent galaxy ZF-COSMOS-20115 (ZF; Glazebrook et al. 2017) can be accommodated in current galaxy formation models. In our model, a population of quiescent galaxies (QGs) with stellar masses and star formation rates comparable to those of ZF naturally emerges at redshifts $z<4$. There are two and five ZF analogues at the redshift $3.86$ and $3.58$ in the Millennium simulation volume, respectively. We demonstrate that, while the $z>3.5$ massive QGs are rare (about 2\% of the galaxies with the similar stellar masses), the existing AGN feedback model implemented in the semi-analytic galaxy formation model can successfully explain the formation of the high-redshift QGs as it does on their lower redshift counterparts.

astro-ph.GA↗