arXiv subjects

Jun Xu

Publications and source records attributed to Jun Xu.

At least 361 records · Page 20Linked to original sources

arXiv2020

Variance Reduction for Deep Q-Learning using Stochastic Recursive Gradient

Deep Q-learning algorithms often suffer from poor gradient estimations with an excessive variance, resulting in unstable training and poor sampling efficiency. Stochastic variance-reduced gradient methods such as SVRG have been applied to reduce the estimation variance (Zhao et al. 2019). However, due to the online instance generation nature of reinforcement learning, directly applying SVRG to deep Q-learning is facing the problem of the inaccurate estimation of the anchor points, which dramatically limits the potentials of SVRG. To address this issue and inspired by the recursive gradient variance reduction algorithm SARAH (Nguyen et al. 2017), this paper proposes to introduce the recursive framework for updating the stochastic gradient estimates in deep Q-learning, achieving a novel algorithm called SRG-DQN. Unlike the SVRG-based algorithms, SRG-DQN designs a recursive update of the stochastic gradient estimate. The parameter update is along an accumulated direction using the past stochastic gradient information, and therefore can get rid of the estimation of the full gradients as the anchors. Additionally, SRG-DQN involves the Adam process for further accelerating the training process. Theoretical analysis and the experimental results on well-known reinforcement learning tasks demonstrate the efficiency and effectiveness of the proposed SRG-DQN algorithm.

cs.LG↗

arXiv2020

Nucleus giant resonances from an improved isospin-dependent Boltzmann-Uehling-Uhlenbeck transport approach

We have studied the isoscalar giant quadruple resonance (ISGQR) and the isovector giant dipole resonance (IVGDR) in $^{208}$Pb based on an improved isospin-dependent Boltzmann-Uehling-Uhlenbeck transport approach using an improved isospin- and momentum-dependent interaction. With the isoscalar nucleon effective mass and the nucleon-nucleon cross section which reproduces respectively the excitation energy and the width of the ISGQR strength function, the slope parameter of the symmetry energy and the neutron-proton effective mass splitting are constrained respectively within $36<L<62$ MeV and $0.08δ<(m_{n0}^*-m_{p0}^*)/m<0.42δ$, by comparing the resulting centroid energy of the IVGDR and the electric dipole polarizability with the experimental data. It is found that nucleon-nucleon collisions have considerable effects on the resulting electric dipole polarizability, which needs to be measured more accurately in order to pin down isovector nuclear interactions.

nucl-th↗

arXiv2020

Learning to Learn with Variational Information Bottleneck for Domain Generalization

Domain generalization models learn to generalize to previously unseen domains, but suffer from prediction uncertainty and domain shift. In this paper, we address both problems. We introduce a probabilistic meta-learning model for domain generalization, in which classifier parameters shared across domains are modeled as distributions. This enables better handling of prediction uncertainty on unseen domains. To deal with domain shift, we learn domain-invariant representations by the proposed principle of meta variational information bottleneck, we call MetaVIB. MetaVIB is derived from novel variational bounds of mutual information, by leveraging the meta-learning setting of domain generalization. Through episodic training, MetaVIB learns to gradually narrow domain gaps to establish domain-invariant representations, while simultaneously maximizing prediction accuracy. We conduct experiments on three benchmarks for cross-domain visual recognition. Comprehensive ablation studies validate the benefits of MetaVIB for domain generalization. The comparison results demonstrate our method outperforms previous approaches consistently.

cs.CV↗

arXiv2020

Auto-captions on GIF: A Large-scale Video-sentence Dataset for Vision-language Pre-training

In this work, we present Auto-captions on GIF, which is a new large-scale pre-training dataset for generic video understanding. All video-sentence pairs are created by automatically extracting and filtering video caption annotations from billions of web pages. Auto-captions on GIF dataset can be utilized to pre-train the generic feature representation or encoder-decoder structure for video captioning, and other downstream tasks (e.g., sentence localization in videos, video question answering, etc.) as well. We present a detailed analysis of Auto-captions on GIF dataset in comparison to existing video-sentence datasets. We also provide an evaluation of a Transformer-based encoder-decoder structure for vision-language pre-training, which is further adapted to video captioning downstream task and yields the compelling generalizability on MSR-VTT. The dataset is available at \url{http://www.auto-video-captions.top/2020/dataset}.

cs.CV↗

arXiv2020

Enhanced yield ratio of light nuclei in heavy ion collisions with a first-order QCD phase transition

Using a transport model that includes a first-order chiral phase transition between the partonic and the hadronic matter, we study the development of density fluctuations in the matter produced in heavy ion collisions as it undergoes the phase transition, and their time evolution in later hadronic stage of the collisions. Using the coalescence model to describe the production of deuterons and tritons from nucleons at the kinetic freeze out, we find that the yield ratio $ N_\text{t}N_\text{p}/ N_\text{d}^2$, where $N_\text{p}$, $N_\text{d}$, and $N_\text{t}$ are, respectively, the proton, deuteron, and triton numbers, is enhanced if the evolution trajectory of the produced matter in the QCD phase diagram passes through the spinodal region of a first-order chiral phase transition.

nucl-th↗

arXiv2020

Noisy-As-Clean: Learning Self-supervised Denoising from the Corrupted Image

Supervised deep networks have achieved promisingperformance on image denoising, by learning image priors andnoise statistics on plenty pairs of noisy and clean images. Unsupervised denoising networks are trained with only noisy images. However, for an unseen corrupted image, both supervised andunsupervised networks ignore either its particular image prior, the noise statistics, or both. That is, the networks learned from external images inherently suffer from a domain gap problem: the image priors and noise statistics are very different between the training and test images. This problem becomes more clear when dealing with the signal dependent realistic noise. To circumvent this problem, in this work, we propose a novel "Noisy-As-Clean" (NAC) strategy of training self-supervised denoising networks. Specifically, the corrupted test image is directly taken as the "clean" target, while the inputs are synthetic images consisted of this corrupted image and a second and similar corruption. A simple but useful observation on our NAC is: as long as the noise is weak, it is feasible to learn a self-supervised network only with the corrupted image, approximating the optimal parameters of a supervised network learned with pairs of noisy and clean images. Experiments on synthetic and realistic noise removal demonstrate that, the DnCNN and ResNet networks trained with our self-supervised NAC strategy achieve comparable or better performance than the original ones and previous supervised/unsupervised/self-supervised networks. The code is publicly available at https://github.com/csjunxu/Noisy-As-Clean.

cs.CV↗

arXiv2020

Conditional Variational Image Deraining

Image deraining is an important yet challenging image processing task. Though deterministic image deraining methods are developed with encouraging performance, they are infeasible to learn flexible representations for probabilistic inference and diverse predictions. Besides, rain intensity varies both in spatial locations and across color channels, making this task more difficult. In this paper, we propose a Conditional Variational Image Deraining (CVID) network for better deraining performance, leveraging the exclusive generative ability of Conditional Variational Auto-Encoder (CVAE) on providing diverse predictions for the rainy image. To perform spatially adaptive deraining, we propose a spatial density estimation (SDE) module to estimate a rain density map for each image. Since rain density varies across different color channels, we also propose a channel-wise (CW) deraining scheme. Experiments on synthesized and real-world datasets show that the proposed CVID network achieves much better performance than previous deterministic methods on image deraining. Extensive ablation studies validate the effectiveness of the proposed SDE module and CW scheme in our CVID network. The code is available at \url{https://github.com/Yingjun-Du/VID}.

cs.CV↗

arXiv2020

SetRank: Learning a Permutation-Invariant Ranking Model for Information Retrieval

In learning-to-rank for information retrieval, a ranking model is automatically learned from the data and then utilized to rank the sets of retrieved documents. Therefore, an ideal ranking model would be a mapping from a document set to a permutation on the set, and should satisfy two critical requirements: (1)~it should have the ability to model cross-document interactions so as to capture local context information in a query; (2)~it should be permutation-invariant, which means that any permutation of the inputted documents would not change the output ranking. Previous studies on learning-to-rank either design uni-variate scoring functions that score each document separately, and thus failed to model the cross-document interactions; or construct multivariate scoring functions that score documents sequentially, which inevitably sacrifice the permutation invariance requirement. In this paper, we propose a neural learning-to-rank model called SetRank which directly learns a permutation-invariant ranking model defined on document sets of any size. SetRank employs a stack of (induced) multi-head self attention blocks as its key component for learning the embeddings for all of the retrieved documents jointly. The self-attention mechanism not only helps SetRank to capture the local context information from cross-document interactions, but also to learn permutation-equivariant representations for the inputted documents, which therefore achieving a permutation-invariant ranking model. Experimental results on three large scale benchmarks showed that the SetRank significantly outperformed the baselines include the traditional learning-to-rank models and state-of-the-art Neural IR models.

cs.IR↗

arXiv2020

Bilateral Attention Network for RGB-D Salient Object Detection

Most existing RGB-D salient object detection (SOD) methods focus on the foreground region when utilizing the depth images. However, the background also provides important information in traditional SOD methods for promising performance. To better explore salient information in both foreground and background regions, this paper proposes a Bilateral Attention Network (BiANet) for the RGB-D SOD task. Specifically, we introduce a Bilateral Attention Module (BAM) with a complementary attention mechanism: foreground-first (FF) attention and background-first (BF) attention. The FF attention focuses on the foreground region with a gradual refinement style, while the BF one recovers potentially useful salient information in the background region. Benefitted from the proposed BAM module, our BiANet can capture more meaningful foreground and background cues, and shift more attention to refining the uncertain details between foreground and background regions. Additionally, we extend our BAM by leveraging the multi-scale techniques for better SOD performance. Extensive experiments on six benchmark datasets demonstrate that our BiANet outperforms other state-of-the-art RGB-D SOD methods in terms of objective metrics and subjective visual comparison. Our BiANet can run up to 80fps on $224\times224$ RGB-D images, with an NVIDIA GeForce RTX 2080Ti GPU. Comprehensive ablation studies also validate our contributions.

cs.CV↗

arXiv2020

X-ray absorption and 9.7 um silicate feature as a probe of AGN torus structure

The dusty torus plays a vital role in unifying active galactic nuclei (AGNs). However, the physical structure of the torus remains largely unclear. Here we present a systematical investigation of the torus mid-infrared (MIR) spectroscopic feature, i.e., the 9.7 um silicate line, of $175$ AGNs selected from the Swift/BAT Spectroscopic Survey (BASS). Our sample is constructed to ensure that each of the $175$ AGNs has Spizter/IRS MIR, optical, and X-ray spectroscopic coverage. Therefore, we can simultaneously measure the silicate strength, optical emission lines, and X-ray properties (e.g., the column density and the intrinsic X-ray luminosity). We show that, consistent with previous works, the silicate strength is weakly correlated with the hydrogen column density ($N_\mathrm{H}^\mathrm{X}$), albeit with large scatters. For X-ray unobscured AGNs, the silicate-strength-derived $V$-band extinction and the broad-H$α$-inferred one are both small; however, for X-ray obscured AGNs, the former is much larger than the latter. In addition, we find that the optical type 1 AGNs with strong X-ray absorption on average show significant silicate absorption, indicating that their X-ray absorption might not be caused by dust-free gas in the broad-line region. Our results suggest that the distribution and structure of the obscuring dusty torus are likely to be very complex. We test our results against the smooth and clumpy torus models and find evidence in favor of the clumpy torus model.

astro-ph.HE↗

arXiv2020

AGNs are not that cool: revisiting the intrinsic AGN far-infrared spectral energy distribution

We investigate the intrinsic spectral energy distribution (SED) of active galactic nuclei (AGNs) at infrared (IR) bands with 42 $z < 0.5$ optically luminous Palomar Green survey quasars through SED decomposition. We decompose the SEDs of the 42 quasars by combining an AGN IR template library Siebenmorgen2015 that covers a wide range of the AGN parameter space with three commonly used galaxy template libraries. We determine the median AGN SED from the best-fitting results. The far-IR (FIR) contribution of our median AGN SED is significantly smaller than that of Symeonidis et al. 2016, but roughly consistent with that of Lyu et al. 2017. The AGN IR SED becomes cooler with increasing bolometric luminosity, which might be due to that more luminous AGNs might have stronger radiative feedback to change torus structures and/or their tori might have higher metallicities. Our conclusions do not depend on the choice of galaxy template libraries. However, since the predicted polycyclic aromatic hydrocarbon (PAH) emission line flux is galaxy template-dependent, cautions should be taken on deriving galaxy FIR contribution from PAH fluxes.

astro-ph.GA↗

arXiv2020

An Empirical Study on Benchmarks of Artificial Software Vulnerabilities

Recently, various techniques (e.g., fuzzing) have been developed for vulnerability detection. To evaluate those techniques, the community has been developing benchmarks of artificial vulnerabilities because of a shortage of ground-truth. However, people have concerns that such vulnerabilities cannot represent reality and may lead to unreliable and misleading results. Unfortunately, there lacks research on handling such concerns. In this work, to understand how close these benchmarks mirror reality, we perform an empirical study on three artificial vulnerability benchmarks - LAVA-M, Rode0day and CGC (2669 bugs) and various real-world memory-corruption vulnerabilities (80 CVEs). Furthermore, we propose a model to depict the properties of memory-corruption vulnerabilities. Following this model, we conduct intensive experiments and data analyses. Our analytic results reveal that while artificial benchmarks attempt to approach the real world, they still significantly differ from reality. Based on the findings, we propose a set of strategies to improve the quality of artificial benchmarks.

cs.CR↗

arXiv2020

STAR: A Structure and Texture Aware Retinex Model

Retinex theory is developed mainly to decompose an image into the illumination and reflectance components by analyzing local image derivatives. In this theory, larger derivatives are attributed to the changes in reflectance, while smaller derivatives are emerged in the smooth illumination. In this paper, we utilize exponentiated local derivatives (with an exponent γ) of an observed image to generate its structure map and texture map. The structure map is produced by been amplified with γ > 1, while the texture map is generated by been shrank with γ < 1. To this end, we design exponential filters for the local derivatives, and present their capability on extracting accurate structure and texture maps, influenced by the choices of exponents γ. The extracted structure and texture maps are employed to regularize the illumination and reflectance components in Retinex decomposition. A novel Structure and Texture Aware Retinex (STAR) model is further proposed for illumination and reflectance decomposition of a single image. We solve the STAR model by an alternating optimization algorithm. Each sub-problem is transformed into a vectorized least squares regression, with closed-form solutions. Comprehensive experiments on commonly tested datasets demonstrate that, the proposed STAR model produce better quantitative and qualitative performance than previous competing methods, on illumination and reflectance decomposition, low-light image enhancement, and color correction. The code is publicly available at https://github.com/csjunxu/STAR.

cs.CV↗

arXiv2020

NLH: A Blind Pixel-level Non-local Method for Real-world Image Denoising

Non-local self similarity (NSS) is a powerful prior of natural images for image denoising. Most of existing denoising methods employ similar patches, which is a patch-level NSS prior. In this paper, we take one step forward by introducing a pixel-level NSS prior, i.e., searching similar pixels across a non-local region. This is motivated by the fact that finding closely similar pixels is more feasible than similar patches in natural images, which can be used to enhance image denoising performance. With the introduced pixel-level NSS prior, we propose an accurate noise level estimation method, and then develop a blind image denoising method based on the lifting Haar transform and Wiener filtering techniques. Experiments on benchmark datasets demonstrate that, the proposed method achieves much better performance than previous non-deep methods, and is still competitive with existing state-of-the-art deep learning based methods on real-world image denoising. The code is publicly available at https://github.com/njusthyk1972/NLH.

cs.CV↗

arXiv2020

Univariate ReLU neural network and its application in nonlinear system identification

ReLU (rectified linear units) neural network has received significant attention since its emergence. In this paper, a univariate ReLU (UReLU) neural network is proposed to both modelling the nonlinear dynamic system and revealing insights about the system. Specifically, the neural network consists of neurons with linear and UReLU activation functions, and the UReLU functions are defined as the ReLU functions respect to each dimension. The UReLU neural network is a single hidden layer neural network, and the structure is relatively simple. The initialization of the neural network employs the decoupling method, which provides a good initialization and some insight into the nonlinear system. Compared with normal ReLU neural network, the number of parameters of UReLU network is less, but it still provide a good approximation of the nonlinear dynamic system. The performance of the UReLU neural network is shown through a Hysteretic benchmark system: the Bouc-Wen system. Simulation results verify the effectiveness of the proposed method.

eess.SY↗

arXiv2020

Properties of strange quark stars with isovector interactions

We study the properties of strange quark stars by employing a 3-flavor Nambu-Jona-Lasinio model with both scalar-isovector and vector-isovector interactions. Using the constraint on the vector-isoscalar interaction strength obtained from the elliptic flow splitting between particles and their antiparticles in relativistic heavy-ion collisions, we investigate the dependence of the properties of strange quark stars on the vector-isovector and the scalar-isovector interactions, and compare the results with the state-of-art astrophysical constraints on the compact star radius and mass as well as its tidal deformability from the GW170817 event. Results from our study reinforce the prospect of using both heavy-ion collisions and astrophysical observations to provide constraints on the isovector coupling strength in quark matter and thus the quark matter equation of state as well as the QCD phase structure at finite isospin chemical potentials.

nucl-th↗

arXiv2020

Privacy for All: Demystify Vulnerability Disparity of Differential Privacy against Membership Inference Attack

Machine learning algorithms, when applied to sensitive data, pose a potential threat to privacy. A growing body of prior work has demonstrated that membership inference attack (MIA) can disclose specific private information in the training data to an attacker. Meanwhile, the algorithmic fairness of machine learning has increasingly caught attention from both academia and industry. Algorithmic fairness ensures that the machine learning models do not discriminate a particular demographic group of individuals (e.g., black and female people). Given that MIA is indeed a learning model, it raises a serious concern if MIA ``fairly'' treats all groups of individuals equally. In other words, whether a particular group is more vulnerable against MIA than the other groups. This paper examines the algorithmic fairness issue in the context of MIA and its defenses. First, for fairness evaluation, it formalizes the notation of vulnerability disparity (VD) to quantify the difference of MIA treatment on different demographic groups. Second, it evaluates VD on four real-world datasets, and shows that VD indeed exists in these datasets. Third, it examines the impacts of differential privacy, as a defense mechanism of MIA, on VD. The results show that although DP brings significant change on VD, it cannot eliminate VD completely. Therefore, fourth, it designs a new mitigation algorithm named FAIRPICK to reduce VD. An extensive set of experimental results demonstrate that FAIRPICK can effectively reduce VD for both with and without the DP deployment.

cs.CR↗

arXiv2020

Joint Inference on Truth/Rumor and Their Sources in Social Networks

In the contemporary era of information explosion, we are often faced with the mixture of massive \emph{truth} (true information) and \emph{rumor} (false information) flooded over social networks. Under such circumstances, it is very essential to infer whether each claim (e.g., news, messages) is a truth or a rumor, and identify their \emph{sources}, i.e., the users who initially spread those claims. While most prior arts have been dedicated to the two tasks respectively, this paper aims to offer the joint inference on truth/rumor and their sources. Our insight is that a joint inference can enhance the mutual performance on both sides. To this end, we propose a framework named SourceCR, which alternates between two modules, i.e., \emph{credibility-reliability training} for truth/rumor inference and \emph{division-querying} for source detection, in an iterative manner. To elaborate, the former module performs a simultaneous estimation of claim credibility and user reliability by virtue of an Expectation Maximization algorithm, which takes the source reliability outputted from the latter module as the initial input. Meanwhile, the latter module divides the network into two different subnetworks labeled via the claim credibility, and in each subnetwork launches source detection by applying querying of theoretical budget guarantee to the users selected via the estimated reliability from the former module. The proposed SourceCR is provably convergent, and algorithmic implementable with reasonable computational complexity. We empirically validate the effectiveness of the proposed framework in both synthetic and real datasets, where the joint inference leads to an up to 35\% accuracy of credibility gain and 29\% source detection rate gain compared with the separate counterparts.

cs.SI↗