Showing 101–150 of 238 results for author: Xiong, L

Classification Auto-Encoder based Detector against Diverse Data Poisoning Attacks

Authors: Fereshteh Razmi, Li Xiong

Abstract: Poisoning attacks are a category of adversarial machine learning threats in which an adversary attempts to subvert the outcome of the machine learning systems by injecting crafted data into training data set, thus increasing the machine learning model's test error. The adversary can tamper with the data feature space, data labels, or both, each leading to a different attack strategy with different… ▽ More Poisoning attacks are a category of adversarial machine learning threats in which an adversary attempts to subvert the outcome of the machine learning systems by injecting crafted data into training data set, thus increasing the machine learning model's test error. The adversary can tamper with the data feature space, data labels, or both, each leading to a different attack strategy with different strengths. Various detection approaches have recently emerged, each focusing on one attack strategy. The Achilles heel of many of these detection approaches is their dependence on having access to a clean, untampered data set. In this paper, we propose CAE, a Classification Auto-Encoder based detector against diverse poisoned data. CAE can detect all forms of poisoning attacks using a combination of reconstruction and classification errors without having any prior knowledge of the attack strategy. We show that an enhanced version of CAE (called CAE+) does not have to employ a clean data set to train the defense model. Our experimental results on three real datasets MNIST, Fashion-MNIST and CIFAR demonstrate that our proposed method can maintain its functionality under up to 30% contaminated data and help the defended SVM classifier to regain its best accuracy. △ Less

Submitted 16 May, 2022; v1 submitted 9 August, 2021; originally announced August 2021.

Comments: This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible

Narrow quadrupolar surface lattice resonances and band reversal in vertical metal-insulator-metal gratings

Authors: Xinyu Fang, Lei Xiong, Jian** Shi, Hongwei Ding, Guangyuan Li

Abstract: We report narrow quadrupolar surface lattice resonances (SLRs) under normal incidence, and the observation, for the first time, of the band reversal effect of SLRs supported by a vertical metal-insulator-metal nanograting, which is embedded in a homogeneous dielectric environment. Simulation results show that under normal incidence, quadrupolar SLR with linewidth of 1~nm and high quality factor of… ▽ More We report narrow quadrupolar surface lattice resonances (SLRs) under normal incidence, and the observation, for the first time, of the band reversal effect of SLRs supported by a vertical metal-insulator-metal nanograting, which is embedded in a homogeneous dielectric environment. Simulation results show that under normal incidence, quadrupolar SLR with linewidth of 1~nm and high quality factor of 979 can be excited in the near-infrared regime, and that under oblique incidence, out-of-plane dipolar SLRs of relatively large quality factors (>=150) can be launched. By varying the incidence angle, the SLR wavelength can be continuously tuned over an extremely broadband range of 750 nm, covering most of the near-infrared regime, and the quality factor decreases exponentially. Remarkably, the resonance lineshape can also be dynamically tuned from an asymmetric Fano-shaped dip to a peak, a dip/peak pair, and a perfect symmetric Lorentzian peak, suggesting the appearance of the band reversal effect. We expect the high-Q SLRs with broadband tunability and tunable lineshapes will find potential applications in enhanced nanoscale light-matter interactions in nanolasers, nonlinear optics and sensing. △ Less

Submitted 5 August, 2021; originally announced August 2021.

Comments: 11 pages, 6 figures

Label Distribution Amendment with Emotional Semantic Correlations for Facial Expression Recognition

Authors: Shasha Mao, Guanghui Shi, Licheng Jiao, Shui** Gou, Yangyang Li, Lin Xiong, Boxin Shi

Abstract: By utilizing label distribution learning, a probability distribution is assigned for a facial image to express a compound emotion, which effectively improves the problem of label uncertainties and noises occurred in one-hot labels. In practice, it is observed that correlations among emotions are inherently different, such as surprised and happy emotions are more possibly synchronized than surprise… ▽ More By utilizing label distribution learning, a probability distribution is assigned for a facial image to express a compound emotion, which effectively improves the problem of label uncertainties and noises occurred in one-hot labels. In practice, it is observed that correlations among emotions are inherently different, such as surprised and happy emotions are more possibly synchronized than surprised and neutral. It indicates the correlation may be crucial for obtaining a reliable label distribution. Based on this, we propose a new method that amends the label distribution of each facial image by leveraging correlations among expressions in the semantic space. Inspired by inherently diverse correlations among word2vecs, the topological information among facial expressions is firstly explored in the semantic space, and each image is embedded into the semantic space. Specially, a class-relation graph is constructed to transfer the semantic correlation among expressions into the task space. By comparing semantic and task class-relation graphs of each image, the confidence of its label distribution is evaluated. Based on the confidence, the label distribution is amended by enhancing samples with higher confidence and weakening samples with lower confidence. Experimental results demonstrate the proposed method is more effective than compared state-of-the-art methods. △ Less

Submitted 23 July, 2021; originally announced July 2021.

RobustFed: A Truth Inference Approach for Robust Federated Learning

Authors: Farnaz Tahmasebian, Jian Lou, Li Xiong

Abstract: Federated learning is a prominent framework that enables clients (e.g., mobile devices or organizations) to train a collaboratively global model under a central server's orchestration while kee** local training datasets' privacy. However, the aggregation step in federated learning is vulnerable to adversarial attacks as the central server cannot manage clients' behavior. Therefore, the global mo… ▽ More Federated learning is a prominent framework that enables clients (e.g., mobile devices or organizations) to train a collaboratively global model under a central server's orchestration while kee** local training datasets' privacy. However, the aggregation step in federated learning is vulnerable to adversarial attacks as the central server cannot manage clients' behavior. Therefore, the global model's performance and convergence of the training process will be affected under such attacks.To mitigate this vulnerability issue, we propose a novel robust aggregation algorithm inspired by the truth inference methods in crowdsourcing via incorporating the worker's reliability into aggregation. We evaluate our solution on three real-world datasets with a variety of machine learning models. Experimental results show that our solution ensures robust federated learning and is resilient to various types of attacks, including noisy data attacks, Byzantine attacks, and label flip** attacks. △ Less

Submitted 18 July, 2021; originally announced July 2021.

Comments: 16 pages

Investor Behavior Modeling by Analyzing Financial Advisor Notes: A Machine Learning Perspective

Authors: Cynthia Pagliaro, Dhagash Mehta, Han-Tai Shiao, Shaofei Wang, Luwei Xiong

Abstract: Modeling investor behavior is crucial to identifying behavioral coaching opportunities for financial advisors. With the help of natural language processing (NLP) we analyze an unstructured (textual) dataset of financial advisors' summary notes, taken after every investor conversation, to gain first ever insights into advisor-investor interactions. These insights are used to predict investor needs… ▽ More Modeling investor behavior is crucial to identifying behavioral coaching opportunities for financial advisors. With the help of natural language processing (NLP) we analyze an unstructured (textual) dataset of financial advisors' summary notes, taken after every investor conversation, to gain first ever insights into advisor-investor interactions. These insights are used to predict investor needs during adverse market conditions; thus allowing advisors to coach investors and help avoid inappropriate financial decision-making. First, we perform topic modeling to gain insight into the emerging topics and trends. Based on this insight, we construct a supervised classification model to predict the probability that an advised investor will require behavioral coaching during volatile market periods. To the best of our knowledge, ours is the first work on exploring the advisor-investor relationship using unstructured data. This work may have far-reaching implications for both traditional and emerging financial advisory service models like robo-advising. △ Less

Submitted 12 July, 2021; originally announced July 2021.

Comments: 8 pages, 2 column format, 7 figures+5 tables

Federated Graph Classification over Non-IID Graphs

Authors: Han Xie, **g Ma, Li Xiong, Carl Yang

Abstract: Federated learning has emerged as an important paradigm for training machine learning models in different domains. For graph-level tasks such as graph classification, graphs can also be regarded as a special type of data samples, which can be collected and stored in separate local systems. Similar to other domains, multiple local systems, each holding a small set of graphs, may benefit from collab… ▽ More Federated learning has emerged as an important paradigm for training machine learning models in different domains. For graph-level tasks such as graph classification, graphs can also be regarded as a special type of data samples, which can be collected and stored in separate local systems. Similar to other domains, multiple local systems, each holding a small set of graphs, may benefit from collaboratively training a powerful graph mining model, such as the popular graph neural networks (GNNs). To provide more motivation towards such endeavors, we analyze real-world graphs from different domains to confirm that they indeed share certain graph properties that are statistically significant compared with random graphs. However, we also find that different sets of graphs, even from the same domain or same dataset, are non-IID regarding both graph structures and node features. To handle this, we propose a graph clustered federated learning (GCFL) framework that dynamically finds clusters of local systems based on the gradients of GNNs, and theoretically justify that such clusters can reduce the structure and feature heterogeneity among graphs owned by the local systems. Moreover, we observe the gradients of GNNs to be rather fluctuating in GCFL which impedes high-quality clustering, and design a gradient sequence-based clustering mechanism based on dynamic time war** (GCFL+). Extensive experimental results and in-depth analysis demonstrate the effectiveness of our proposed frameworks. △ Less

Submitted 7 November, 2021; v1 submitted 25 June, 2021; originally announced June 2021.

Comments: Accepted to NeurIPS 2021

PAM: Understanding Product Images in Cross Product Category Attribute Extraction

Authors: Rongmei Lin, Xiang He, Jie Feng, Nasser Zalmout, Yan Liang, Li Xiong, Xin Luna Dong

Abstract: Understanding product attributes plays an important role in improving online shop** experience for customers and serves as an integral part for constructing a product knowledge graph. Most existing methods focus on attribute extraction from text description or utilize visual information from product images such as shape and color. Compared to the inputs considered in prior works, a product image… ▽ More Understanding product attributes plays an important role in improving online shop** experience for customers and serves as an integral part for constructing a product knowledge graph. Most existing methods focus on attribute extraction from text description or utilize visual information from product images such as shape and color. Compared to the inputs considered in prior works, a product image in fact contains more information, represented by a rich mixture of words and visual clues with a layout carefully designed to impress customers. This work proposes a more inclusive framework that fully utilizes these different modalities for attribute extraction. Inspired by recent works in visual question answering, we use a transformer based sequence to sequence model to fuse representations of product text, Optical Character Recognition (OCR) tokens and visual objects detected in the product image. The framework is further extended with the capability to extract attribute value across multiple product categories with a single model, by training the decoder to predict both product category and attribute value and conditioning its output on product category. The model provides a unified attribute extraction solution desirable at an e-commerce platform that offers numerous product categories with a diverse body of product attributes. We evaluated the model on two product attributes, one with many possible values and one with a small set of possible values, over 14 product categories and found the model could achieve 15% gain on the Recall and 10% gain on the F1 score compared to existing methods using text-only features. △ Less

Submitted 8 June, 2021; originally announced June 2021.

Comments: KDD 2021

Enhanced superconductivity in bilayer PtTe$_2$ by alkali-metal intercalations

Authors: Danhong Wu, Yi** Lin, Lingxiao Xiong, Junjie Li, Tiantian Luo, Deyi Chen, Feipeng Zheng

Abstract: Layered platinum tellurium (PtTe2) was recently synthesized with controllable layer numbers down to a monolayer limit. Using ab initio calculations based on anisotropic Midgal-Eliashberg formalism, we show that by rubidium (Rb) intercalation, weak superconductivity in bilayer PtTe2 can be significantly boosted with superconducting Tc = 8 K in the presence of spin-orbit coupling (SOC). The intercal… ▽ More Layered platinum tellurium (PtTe2) was recently synthesized with controllable layer numbers down to a monolayer limit. Using ab initio calculations based on anisotropic Midgal-Eliashberg formalism, we show that by rubidium (Rb) intercalation, weak superconductivity in bilayer PtTe2 can be significantly boosted with superconducting Tc = 8 K in the presence of spin-orbit coupling (SOC). The intercalant on one hand mediates the interlayer coupling and serves as an electron donor, leading to large density of states at Fermi energy. On the other hand, it increases the mass-enhancement parameter with electron-phonon coupling strength comparable to that of Pt. The potassium intercalated bilayer PtTe2 has a comparable Tc to the case of Rb intercalation. The relatively high Tc with SOC combined with experimental accessible crystal structures suggest that these superconductors are promising platforms to study the novel quantum physics associated with two-dimensional superconductivity, such as the recently proposed type-II Ising superconductivity. △ Less

Submitted 29 May, 2021; originally announced May 2021.

Comments: 7 pafges, 4 figures. Accepted by PRB(2021)

View Distillation with Unlabeled Data for Extracting Adverse Drug Effects from User-Generated Data

Authors: Payam Karisani, **ho D. Choi, Li Xiong

Abstract: We present an algorithm based on multi-layer transformers for identifying Adverse Drug Reactions (ADR) in social media data. Our model relies on the properties of the problem and the characteristics of contextual word embeddings to extract two views from documents. Then a classifier is trained on each view to label a set of unlabeled documents to be used as an initializer for a new classifier in t… ▽ More We present an algorithm based on multi-layer transformers for identifying Adverse Drug Reactions (ADR) in social media data. Our model relies on the properties of the problem and the characteristics of contextual word embeddings to extract two views from documents. Then a classifier is trained on each view to label a set of unlabeled documents to be used as an initializer for a new classifier in the other view. Finally, the initialized classifier in each view is further trained using the initial training examples. We evaluated our model in the largest publicly available ADR dataset. The experiments testify that our model significantly outperforms the transformer-based models pretrained on domain-specific data. △ Less

Submitted 24 May, 2021; originally announced May 2021.

Comments: NAACL 2021 (workshops)

A General Principle: π-Conjugated Confinement Maximizes Band Gap of DUV NLO Materials

Authors: Lin Xiong, Li-Ming Wu, Ling Chen

Abstract: Current nonlinear optical materials face a conventional limitation on the tradeoff between band gap and birefringence, especially in the deep UV spectral region. To circumvent such a dilemma, we propose a general principle, a π-conjugated confinement, to partially decouple the inter group π-conjugated interactions with the separation of a non-π-conjugated group so as to maximize the band gap in co… ▽ More Current nonlinear optical materials face a conventional limitation on the tradeoff between band gap and birefringence, especially in the deep UV spectral region. To circumvent such a dilemma, we propose a general principle, a π-conjugated confinement, to partially decouple the inter group π-conjugated interactions with the separation of a non-π-conjugated group so as to maximize the band gap in comparison with those of simple π-conjugated salts, such as borates, carbonates. Meanwhile, to maintain a large optical anisotropy. We uncover that the π-conjugated confinement is a shared structural feature for all the known DUV NLO materials with favorable properties (45 compounds), and thus, it provides an essential design criterion. Guided by this principle, the carbonophosphate is predicted theoretically for the first time as a promising DUV candidate system, Sr3Y[PO4][CO3]3 and Na3X[PO4][CO3] (X = Ba, Sr, Ca, Mg) exhibit an enhanced birefringence that is 3-24 times larger than that of the simple phosphate, as well as an increased band gap that is 0.2-1.7 eV wider than that of the simple carbonate. Especially, the shortest SHG output of Sr3Y[PO4][CO3]3 is at λPM = 181 nm, being the shortest one among phosphates to date. △ Less

Submitted 23 April, 2021; originally announced April 2021.

CrowdTeacher: Robust Co-teaching with Noisy Answers & Sample-specific Perturbations for Tabular Data

Authors: Mani Sotoodeh, Li Xiong, Joyce C. Ho

Abstract: Samples with ground truth labels may not always be available in numerous domains. While learning from crowdsourcing labels has been explored, existing models can still fail in the presence of sparse, unreliable, or diverging annotations. Co-teaching methods have shown promising improvements for computer vision problems with noisy labels by employing two classifiers trained on each others' confiden… ▽ More Samples with ground truth labels may not always be available in numerous domains. While learning from crowdsourcing labels has been explored, existing models can still fail in the presence of sparse, unreliable, or diverging annotations. Co-teaching methods have shown promising improvements for computer vision problems with noisy labels by employing two classifiers trained on each others' confident samples in each batch. Inspired by the idea of separating confident and uncertain samples during the training process, we extend it for the crowdsourcing problem. Our model, CrowdTeacher, uses the idea that perturbation in the input space model can improve the robustness of the classifier for noisy labels. Treating crowdsourcing annotations as a source of noisy labeling, we perturb samples based on the certainty from the aggregated annotations. The perturbed samples are fed to a Co-teaching algorithm tuned to also accommodate smaller tabular data. We showcase the boost in predictive power attained using CrowdTeacher for both synthetic and real datasets across various label density settings. Our experiments reveal that our proposed approach beats baselines modeling individual annotations and then combining them, methods simultaneously learning a classifier and inferring truth labels, and the Co-teaching algorithm with aggregated labels through common truth inference methods. △ Less

Submitted 31 March, 2021; originally announced March 2021.

MonoRUn: Monocular 3D Object Detection by Reconstruction and Uncertainty Propagation

Authors: Hansheng Chen, Yuyao Huang, Wei Tian, Zhong Gao, Lu Xiong

Abstract: Object localization in 3D space is a challenging aspect in monocular 3D object detection. Recent advances in 6DoF pose estimation have shown that predicting dense 2D-3D correspondence maps between image and object 3D model and then estimating object pose via Perspective-n-Point (PnP) algorithm can achieve remarkable localization accuracy. Yet these methods rely on training with ground truth of obj… ▽ More Object localization in 3D space is a challenging aspect in monocular 3D object detection. Recent advances in 6DoF pose estimation have shown that predicting dense 2D-3D correspondence maps between image and object 3D model and then estimating object pose via Perspective-n-Point (PnP) algorithm can achieve remarkable localization accuracy. Yet these methods rely on training with ground truth of object geometry, which is difficult to acquire in real outdoor scenes. To address this issue, we propose MonoRUn, a novel detection framework that learns dense correspondences and geometry in a self-supervised manner, with simple 3D bounding box annotations. To regress the pixel-related 3D object coordinates, we employ a regional reconstruction network with uncertainty awareness. For self-supervised training, the predicted 3D coordinates are projected back to the image plane. A Robust KL loss is proposed to minimize the uncertainty-weighted reprojection error. During testing phase, we exploit the network uncertainty by propagating it through all downstream modules. More specifically, the uncertainty-driven PnP algorithm is leveraged to estimate object pose and its covariance. Extensive experiments demonstrate that our proposed approach outperforms current state-of-the-art methods on KITTI benchmark. △ Less

Submitted 24 March, 2021; v1 submitted 23 March, 2021; originally announced March 2021.

Comments: CVPR 2021

Fizeau Drag in Graphene Plasmonics

Authors: Y. Dong, L. Xiong, I. Y. Phinney, Z. Sun, R. **g, A. S. McLeod, S. Zhang, S. Liu, F. L. Ruta, H. Gao, Z. Dong, R. Pan, J. H. Edgar, P. Jarillo-Herrero, L. S. Levitov, A. J. Millis, M. M. Fogler, D. A. Bandurin, D. N. Basov

Abstract: Dragging of light by moving dielectrics was predicted by Fresnel and verified by Fizeau's celebrated experiments with flowing water. This momentous discovery is among the experimental cornerstones of Einstein's special relativity and is well understood in the context of relativistic kinematics. In contrast, experiments on dragging photons by an electron flow in solids are riddled with inconsistenc… ▽ More Dragging of light by moving dielectrics was predicted by Fresnel and verified by Fizeau's celebrated experiments with flowing water. This momentous discovery is among the experimental cornerstones of Einstein's special relativity and is well understood in the context of relativistic kinematics. In contrast, experiments on dragging photons by an electron flow in solids are riddled with inconsistencies and so far eluded agreement with the theory. Here we report on the electron flow dragging surface plasmon polaritons (SPPs): hybrid quasiparticles of infrared photons and electrons in graphene. The drag is visualized directly through infrared nano-imaging of propagating plasmonic waves in the presence of a high-density current. The polaritons in graphene shorten their wavelength when launched against the drifting carriers. Unlike the Fizeau effect for light, the SPP drag by electrical currents defies the simple kinematics interpretation and is linked to the nonlinear electrodynamics of the Dirac electrons in graphene. The observed plasmonic Fizeau drag enables breaking of time-reversal symmetry and reciprocity at infrared frequencies without resorting to magnetic fields or chiral optical pum**. △ Less

Submitted 19 March, 2021; originally announced March 2021.

Comments: 11 pages, 3 figures

Journal ref: Nature volume 594, pages513-516 (2021)

Exciton emissions in bilayer WSe2 tuned by the ferroelectric polymer

Authors: Sixin Zhu, Dan Li, Jianlu Wang, Qiang Wang, Yongpeng Wu, Liang Xiong, Zhangfeng Jiang, Huihong Lin, Zhirui Gong, Qi Qin, Xingjun Wang

Abstract: In this work, a hybrid integration of few-layer transition metal dichalcogenides (TMDCs) and ferroelectric polymer is designed to achieve passive control of optical properties in-situ. The electrical polarization in ferroelectric P(VDF-TrFE) polymer can regulate the photoluminescence (PL) in few-layer TMDCs. The total PL intensity is substantially suppressed or enhanced under opposite polarization… ▽ More In this work, a hybrid integration of few-layer transition metal dichalcogenides (TMDCs) and ferroelectric polymer is designed to achieve passive control of optical properties in-situ. The electrical polarization in ferroelectric P(VDF-TrFE) polymer can regulate the photoluminescence (PL) in few-layer TMDCs. The total PL intensity is substantially suppressed or enhanced under opposite polarization in bilayer WSe2. This is because electrons transfer between valley K and Λ in the conduction band induced by the built-in electric field in P(VDF-TrFE) polymer. This charge transfer further changes the competing dynamics between direct and indirect exciton recombination path and overall optical radiation efficiency. We also illustrate that the engineered PL originates from external electric field dependent transferred electron effect. The theoretical result matches the experimental data well. This work demonstrates a device platform in which passive regulation is achieved using 2D TMDCs modulated by polarized ferroelectric materials. △ Less

Submitted 12 April, 2021; v1 submitted 3 March, 2021; originally announced March 2021.

Learning with Hyperspherical Uniformity

Authors: Weiyang Liu, Rongmei Lin, Zhen Liu, Li Xiong, Bernhard Schölkopf, Adrian Weller

Abstract: Due to the over-parameterization nature, neural networks are a powerful tool for nonlinear function approximation. In order to achieve good generalization on unseen data, a suitable inductive bias is of great importance for neural networks. One of the most straightforward ways is to regularize the neural network with some additional objectives. L2 regularization serves as a standard regularization… ▽ More Due to the over-parameterization nature, neural networks are a powerful tool for nonlinear function approximation. In order to achieve good generalization on unseen data, a suitable inductive bias is of great importance for neural networks. One of the most straightforward ways is to regularize the neural network with some additional objectives. L2 regularization serves as a standard regularization for neural networks. Despite its popularity, it essentially regularizes one dimension of the individual neuron, which is not strong enough to control the capacity of highly over-parameterized neural networks. Motivated by this, hyperspherical uniformity is proposed as a novel family of relational regularizations that impact the interaction among neurons. We consider several geometrically distinct ways to achieve hyperspherical uniformity. The effectiveness of hyperspherical uniformity is justified by theoretical insights and empirical evaluations. △ Less

Submitted 16 November, 2021; v1 submitted 2 March, 2021; originally announced March 2021.

Comments: AISTATS 2021 (v3: fixed typos)

Accretion Torque Reversals in GRO J1008-57 Revealed by Insight-HXMT

Authors: W. Wang, Y. M. Tang, Y. L. Tuo, P. R. Epili, S. N. Zhang, L. M. Song, F. J. Lu, J. L. Qu, S. Zhang, M. Y. Ge, Y. Huang, B. Li, Q. C. Bu, C. Cai, X. L. Cao, Z. Chang, L. Chen, T. X. Chen, Y. B. Chen, Y. Chen, Y. P. Chen, W. W. Cui, Y. Y. Du, G. H. Gao, H. Gao , et al. (70 additional authors not shown)

Abstract: GRO J1008-57, as a Be/X-ray transient pulsar, is considered to have the highest magnetic field in known neutron star X-ray binary systems. Observational data of the X-ray outbursts in GRO J1008-57 from 2017 to 2020 were collected by the Insight-HXMT satellite. In this work, the spin period of the neutron star in GRO J1008-57 was determined to be about 93.28 seconds in August 2017, 93.22 seconds in… ▽ More GRO J1008-57, as a Be/X-ray transient pulsar, is considered to have the highest magnetic field in known neutron star X-ray binary systems. Observational data of the X-ray outbursts in GRO J1008-57 from 2017 to 2020 were collected by the Insight-HXMT satellite. In this work, the spin period of the neutron star in GRO J1008-57 was determined to be about 93.28 seconds in August 2017, 93.22 seconds in February 2018, 93.25 seconds in June 2019 and 93.14 seconds in June 2020. GRO J1008-57 evolved in the spin-up process with a mean rate of $-(2.10\pm 0.05)\times$10$^{-4}$ s/d from 2009 -- 2018, and turned into a spin down process with a rate of $(6.7\pm 0.6)\times$10$^{-5}$ s/d from Feb 2018 to June 2019. During the type II outburst of 2020, GRO J1008-57 had the spin-up torque again. During the torque reversals, the pulse profiles and continuum X-ray spectra did not change significantly, and the cyclotron resonant scattering feature around 80 keV was only detected during the outbursts in 2017 and 2020. Based on the observed mean spin-up rate, we estimated the inner accretion disk radius in GRO J1008-57 (about 1 - 2 times of the Alfvén radius) by comparing different accretion torque models of magnetic neutron stars. During the spin-down process, the magnetic torque should dominate over the matter accreting inflow torque, and we constrained the surface dipole magnetic field $B\geq 6\times 10^{12}$ G for the neutron star in GRO J1008-57, which is consistent with the magnetic field strength obtained by cyclotron line centroid energy. △ Less

Submitted 24 February, 2021; originally announced February 2021.

Comments: 10 pages, 5 figures, and 3 tables, the Journal of High Energy Astrophysics in press

QPOs and Orbital elements of X-ray binary 4U 0115+63 during the 2017 outburst observed by Insight-HXMT

Authors: Y. Z. Ding, W. Wang, P. Zhang, Q. C. Bu, C. Cai, X. L. Cao, C. Zhi, L. Chen, T. X. Chen, Y. B. Chen, Y. Chen, Y. P. Chen, W. W. Cui, Y. Y. Du, G. H. Gao, H. Gao, M. Y. Ge, Y. D. Gu, J. Guan, C. C. Guo, D. W. Han, Y. Huang, J. Huo, S. M. Jia, W. C. Jiang , et al. (69 additional authors not shown)

Abstract: In this paper, we presented a detailed timing analysis of a prominent outburst of 4U 0115+63 detected by \textit{Insight}-HXMT in 2017 August. The spin period of the neutron star was determined to be $3.61398\pm 0.00002$ s at MJD 57978. We measured the period variability and extract the orbital elements of the binary system. The angle of periastron evolved with a rate of $0.048\pm0.003$ $yr^{-1}$.… ▽ More In this paper, we presented a detailed timing analysis of a prominent outburst of 4U 0115+63 detected by \textit{Insight}-HXMT in 2017 August. The spin period of the neutron star was determined to be $3.61398\pm 0.00002$ s at MJD 57978. We measured the period variability and extract the orbital elements of the binary system. The angle of periastron evolved with a rate of $0.048\pm0.003$ $yr^{-1}$. The light curves are folded to sketch the pulse profiles in different energy ranges. A multi-peak structure in 1-10 keV is clearly illustrated. We introduced wavelet analysis into our data analysis procedures to study QPO signals and perform a detailed wavelet analysis in many different energy ranges. Through the wavelet spectra, we report the discovery of a QPO at the frequency $\sim 10$ mHz. In addition, the X-ray light curves showed multiple QPOs in the period of $\sim 16-32 $ s and $\sim 67- 200 $ s. We found that the $\sim100$ s QPO was significant in most of the observations and energies. There exist positive relations between X-ray luminosity and their Q-factors and S-factors, while the QPO periods have no correlation with X-ray luminosity. In wavelet phase maps, we found that the pulse phase of $\sim 67- 200 $ s QPO drifting frequently while the $\sim 16-32 $ s QPO scarcely drifting. The dissipation of oscillations from high energy to low energy was also observed. These features of QPOs in 4U 0115+63 provide new challenge to our understanding of their physical origins. △ Less

Submitted 18 February, 2021; originally announced February 2021.

Comments: 14 pages, 9 figures, and 6 tables. This work has been submitted to MNRAS after the referee's report

The Incompressible Navier-Stokes-Fourier Limit from Boltzmann-Fermi-Dirac Equation

Authors: Ning Jiang, Linjie Xiong, Kai Zhou

Abstract: We study Boltzmann-Fermi-Dirac equation when quantum effects are taken into account in dilute gas dynamics. By employing new estimates on trilinear terms of collision kernels, we prove the global existence of the classical solution to Boltzmann-Fermi-Dirac equation near equilibrium. Furthermore, the limit from Boltzmann-Fermi-Dirac equation to incompressible Navier-Stokes-Fourier equations is just… ▽ More We study Boltzmann-Fermi-Dirac equation when quantum effects are taken into account in dilute gas dynamics. By employing new estimates on trilinear terms of collision kernels, we prove the global existence of the classical solution to Boltzmann-Fermi-Dirac equation near equilibrium. Furthermore, the limit from Boltzmann-Fermi-Dirac equation to incompressible Navier-Stokes-Fourier equations is justified rigorously. The corresponding formal analysis was given in the thesis of Zakrevskiy \cite{Zakrevskiy} △ Less

Submitted 4 February, 2021; originally announced February 2021.

Comments: 32 pages

An Analysis Of Protected Health Information Leakage In Deep-Learning Based De-Identification Algorithms

Authors: Salman Seyedi, Li Xiong, Shamim Nemati, Gari D. Clifford

Abstract: The increasing complexity of algorithms for analyzing medical data, including de-identification tasks, raises the possibility that complex algorithms are learning not just the general representation of the problem, but specifics of given individuals within the data. Modern legal frameworks specifically prohibit the intentional or accidental distribution of patient data, but have not addressed this… ▽ More The increasing complexity of algorithms for analyzing medical data, including de-identification tasks, raises the possibility that complex algorithms are learning not just the general representation of the problem, but specifics of given individuals within the data. Modern legal frameworks specifically prohibit the intentional or accidental distribution of patient data, but have not addressed this potential avenue for leakage of such protected health information. Modern deep learning algorithms have the highest potential of such leakage due to complexity of the models. Recent research in the field has highlighted such issues in non-medical data, but all analysis is likely to be data and algorithm specific. We, therefore, chose to analyze a state-of-the-art free-text de-identification algorithm based on LSTM (Long Short-Term Memory) and its potential in encoding any individual in the training set. Using the i2b2 Challenge Data, we trained, then analyzed the model to assess whether the output of the LSTM, before the compression layer of the classifier, could be used to estimate the membership of the training data. Furthermore, we used different attacks including membership inference attack method to attack the model. Results indicate that the attacks could not identify whether members of the training data were distinguishable from non-members based on the model output. This indicates that the model does not provide any strong evidence into the identification of the individuals in the training data set and there is not yet empirical evidence it is unsafe to distribute the model for general use. △ Less

Submitted 21 May, 2021; v1 submitted 28 January, 2021; originally announced January 2021.

Comments: 8 pages, 5 figures, PPAI-2021

Transparent Contribution Evaluation for Secure Federated Learning on Blockchain

Authors: Shuaicheng Ma, Yang Cao, Li Xiong

Abstract: Federated Learning is a promising machine learning paradigm when multiple parties collaborate to build a high-quality machine learning model. Nonetheless, these parties are only willing to participate when given enough incentives, such as a fair reward based on their contributions. Many studies explored Shapley value based methods to evaluate each party's contribution to the learned model. However… ▽ More Federated Learning is a promising machine learning paradigm when multiple parties collaborate to build a high-quality machine learning model. Nonetheless, these parties are only willing to participate when given enough incentives, such as a fair reward based on their contributions. Many studies explored Shapley value based methods to evaluate each party's contribution to the learned model. However, they commonly assume a semi-trusted server to train the model and evaluate the data owners' model contributions, which lacks transparency and may hinder the success of federated learning in practice. In this work, we propose a blockchain-based federated learning framework and a protocol to transparently evaluate each participant's contribution. Our framework protects all parties' privacy in the model building phase and transparently evaluates contributions based on the model updates. The experiment with the handwritten digits dataset demonstrates that the proposed method can effectively evaluate the contributions. △ Less

Submitted 24 April, 2021; v1 submitted 26 January, 2021; originally announced January 2021.

Journal ref: 2021 IEEE 37th International Conference on Data Engineering Workshops (ICDEW)

Highly Distorted Lattices in Chemically Complex Alloys Produce Ultra-Elastic Materials with Extraordinary Elinvar Effects

Authors: Q. F. He, J. G. Wang, H. A. Chen, Z. Y. Ding, Z. Q. Zhou, L. H. Xiong, J. H. Luan, J. M. Pelletier, J. C. Qiao, Q. Wang, L. L. Fan, Y. Ren, Q. S. Zeng, C. T. Liu, C. W. Pao, D. J. Srolovitz, Y. Yang

Abstract: Conventional crystalline alloys usually possess a low atomic size difference in order to stabilize its crystalline structure. However, in this article, we report a single phase chemically complex alloy which possesses a large atomic size misfit usually unaffordable to conventional alloys. Consequently, this alloy develops a rather complex atomic-scale chemical order and a highly distorted crystall… ▽ More Conventional crystalline alloys usually possess a low atomic size difference in order to stabilize its crystalline structure. However, in this article, we report a single phase chemically complex alloy which possesses a large atomic size misfit usually unaffordable to conventional alloys. Consequently, this alloy develops a rather complex atomic-scale chemical order and a highly distorted crystalline structure. As a result, this crystalline alloy displays an unusually high elastic strain limit (~2%), about ten times of that of conventional alloys, and an extremely low internal friction (< 2E-4) at room temperature. More interestingly, this alloy firmly maintains its elastic modulus even when the testing temperature rises from room temperature to 900 K, which is unmatched by the existing alloys hitherto reported. From an application viewpoint, our discovery may open up new opportunities to design high precision devices usable even under an extreme environment. △ Less

Submitted 7 January, 2021; originally announced January 2021.

On traceable iterated line graph and hamiltonian path index

Authors: Zhaohong Nou, Liming Xiong, Weihua Yang

Abstract: Xiong and Liu [L. Xiong and Z. Liu, Hamiltonian iterated line graphs, Discrete Math. 256 (2002) 407-422] gave a characterization of the graphs $G$ for which the $n$-th iterated line graph $L^n(G)$ is hamiltonian, for $n\ge2$. In this paper, we study the existence of a hamiltonian path in $L^n(G)$, and give a characterization of $G$ for which $L^n(G)$ has a hamiltonian path. As applications, we use… ▽ More Xiong and Liu [L. Xiong and Z. Liu, Hamiltonian iterated line graphs, Discrete Math. 256 (2002) 407-422] gave a characterization of the graphs $G$ for which the $n$-th iterated line graph $L^n(G)$ is hamiltonian, for $n\ge2$. In this paper, we study the existence of a hamiltonian path in $L^n(G)$, and give a characterization of $G$ for which $L^n(G)$ has a hamiltonian path. As applications, we use this characterization to give several upper bounds on the hamiltonian path index of a graph. △ Less

Submitted 27 December, 2020; originally announced December 2020.

Comments: 12pages

Insight-HXMT observations of Swift J0243.6+6124: the evolution of RMS pulse fractions at super-Eddington luminosity

Authors: P. J. Wang, L. D. Kong, S. Zhang, Y. P. Chen, S. N. Zhang, J. L. Qu, L. Ji, L. Tao, M. Y. Ge, F. J. Lu, L. Chen, L. M. Song, T. P. Li, Y. P. Xu, X. L. Cao, Y. Chen, C. Z. Liu, Q. C. Bu, C. Cai, Z. Chang, G. Chen, T. X. Chen, Y. B. Chen, W. Cui, W. W. Cui , et al. (95 additional authors not shown)

Abstract: Based on Insight-HXMT data, we report on the pulse fraction evolution during the 2017-2018 outburst of the newly discovered first Galactic ultraluminous X-ray source (ULX) Swift J0243.6+6124. The pulse fractions of 19 observation pairs selected in the rising and fading phases with similar luminosity are investigated. The results show a general trend of the pulse fraction increasing with luminosity… ▽ More Based on Insight-HXMT data, we report on the pulse fraction evolution during the 2017-2018 outburst of the newly discovered first Galactic ultraluminous X-ray source (ULX) Swift J0243.6+6124. The pulse fractions of 19 observation pairs selected in the rising and fading phases with similar luminosity are investigated. The results show a general trend of the pulse fraction increasing with luminosity and energy at super-critical luminosity. However, the relative strength of the pulsation between each pair evolves strongly with luminosity. The pulse fraction in the rising phase is larger at luminosity below $7.71\times10^{38}$~erg~s$^{-1}$, but smaller at above. A transition luminosity is found to be energy independent. Such a phenomena is firstly confirmed by Insight-HXMT observations and we speculate it may have relation with the radiation pressure dominated accretion disk. △ Less

Submitted 24 December, 2020; originally announced December 2020.

Physical origin of the nonphysical spin evolution of MAXI J1820+070

Authors: J. Guan, L. Tao, J. L. Qu, S. N. Zhang, W. Zhang, S. Zhang, R. C. Ma, M. Y. Ge, L. M. Song, F. J. Lu, T. P. Li, Y. P. Xu, Y. Chen, X. L. Cao, C. Z. Liu, L. Zhang, Y. N. Wang, Y. P. Chen, Q. C. Bu, C. Cai, Z. Chang, L. Chen, T. X. Chen, Y. B. Chen, W. W. Cui , et al. (70 additional authors not shown)

Abstract: We report on the Insight-HXMT observations of the new black hole X-ray binary MAXI J1820+070 during its 2018 outburst. Detailed spectral analysis via the continuum fitting method shows an evolution of the inferred spin during its high soft sate. Moreover, the hardness ratio, the non-thermal luminosity and the reflection fraction also undergo an evolution, exactly coincident to the period when the… ▽ More We report on the Insight-HXMT observations of the new black hole X-ray binary MAXI J1820+070 during its 2018 outburst. Detailed spectral analysis via the continuum fitting method shows an evolution of the inferred spin during its high soft sate. Moreover, the hardness ratio, the non-thermal luminosity and the reflection fraction also undergo an evolution, exactly coincident to the period when the inferred spin transition takes place. The unphysical evolution of the spin is attributed to the evolution of the inner disc, which is caused by the collapse of a hot corona due to condensation mechanism or may be related to the deceleration of a jet-like corona. The studies of the inner disc radius and the relation between the disc luminosity and the inner disc radius suggest that, only at a particular epoch, did the inner edge of the disc reach the innermost stable circular orbit and the spin measurement is reliable. We then constrain the spin of MAXI J1820+070 to be a*=0.2^{+0.2}_{-0.3}. Such a slowly spinning black hole possessing a strong jet suggests that its jet activity is driven mainly by the accretion disc rather than by the black hole spin. △ Less

Submitted 31 March, 2021; v1 submitted 22 December, 2020; originally announced December 2020.

Comments: 14 pages, 13 figures, 5 tables, accepted for publication in MNRAS

MSC Class: 83C57

X-ray reprocessing in accreting pulsar GX 301-2 observed with Insight-HXMT

Authors: L. Ji, V. Doroshenko, V. Suleimanov, A. Santangelo, M. Orlandini, J. Liu, L. Ducci, S. N. Zhang, A. Nabizadeh, D. Gavran, S. Zhang, M. Y. Ge, X. B. Li, L. Tao, Q. C. Bu, J. L. Qu, F. J. Lu, L. Chen, L. M. Song, T. P. Li, Y. P. Xu, X. L. Cao, Y. Chen, C. Z. Liu, C. Cai , et al. (78 additional authors not shown)

Abstract: We investigate the absorption and emission features in observations of GX 301-2 detected with Insight-HXMT/LE in 2017-2019. At different orbital phases, we found prominent Fe Kalpha, Kbeta and Ni Kalpha lines, as well as Compton shoulders and Fe K-shell absorption edges. These features are due to the X-ray reprocessing caused by the interaction between the radiation from the source and surrounding… ▽ More We investigate the absorption and emission features in observations of GX 301-2 detected with Insight-HXMT/LE in 2017-2019. At different orbital phases, we found prominent Fe Kalpha, Kbeta and Ni Kalpha lines, as well as Compton shoulders and Fe K-shell absorption edges. These features are due to the X-ray reprocessing caused by the interaction between the radiation from the source and surrounding accretion material. According to the ratio of iron lines Kalpha and Kbeta, we infer the accretion material is in a low ionisation state. We find an orbital-dependent local absorption column density, which has a large value and strong variability around the periastron. We explain its variability as a result of inhomogeneities of the accretion environment and/or instabilities of accretion processes. In addition, the variable local column density is correlated with the equivalent width of the iron Kalpha lines throughout the orbit, which suggests that the accretion material near the neutron star is spherically distributed. △ Less

Submitted 4 December, 2020; originally announced December 2020.

Comments: 10 pages, 5 figures, 2 tables, accepted for publication in MNRAS

Insight-HXMT observations of a possible fast transition from jet to wind dominated state during a huge flare of GRS~1915+105

Authors: L. D. Kong, S. Zhang, Y. P. Chen, S. N. Zhang, L. Ji, P. J. Wang, L. Tao, M. Y. Ge, C. Z. Liu, L. M. Song, F. J. Lu, J. L. Qu, T. P. Li, Y. P. Xu, X. L. Cao, Y. Chen, Q. C. Bu, C. Cai, Z. Chang, G. Chen, L. Chen, T. X. Chen, W. W. Cui, Y. Y. Du, G. H. Gao , et al. (71 additional authors not shown)

Abstract: We present the analysis of the brightest flare that was recorded in the \emph{Insight}-HMXT data set, in a broad energy range (2$-$200 keV) from the microquasar GRS~1915+105 during an unusual low-luminosity state. This flare was detected by \emph{Insight}-HXMT among a series of flares during 2 June 2019 UTC 16:37:06 to 20:11:36, with a 2-200 keV luminosity of 3.4$-$7.27$\times10^{38}$ erg s… ▽ More We present the analysis of the brightest flare that was recorded in the \emph{Insight}-HMXT data set, in a broad energy range (2$-$200 keV) from the microquasar GRS~1915+105 during an unusual low-luminosity state. This flare was detected by \emph{Insight}-HXMT among a series of flares during 2 June 2019 UTC 16:37:06 to 20:11:36, with a 2-200 keV luminosity of 3.4$-$7.27$\times10^{38}$ erg s$^{-1}$. Basing on the broad-band spectral analysis, we find that the flare spectrum shows different behaviors during bright and faint epochs. The spectrum of the flare can be fitted with a model dominated by a power-law component. Additional components show up in the bright epoch with a hard tail and in the faint epoch with an absorption line $\sim$ 6.78 keV. The reflection component of the latter is consistent with an inner disk radius $\sim$ 5 times larger than that of the former. These results on the giant flare during the "unusual" low-luminosity state of GRS~1915+105 may suggest that the source experiences a possible fast transition from a jet-dominated state to a wind-dominated state. We speculate that the evolving accretion disk and the large-scale magnetic field may play important roles in this peculiar huge flare. △ Less

Submitted 4 December, 2020; originally announced December 2020.

Turbulence and heat transfer on a rotating, heated half soap bubble

Authors: X. Q. HE, A. D. Bragg, Y. L. Xiong, P. Fischer

Abstract: We use Direct Numerical Simulations to study the two-dimensional flow of a rotating, half soap bubble that is heated at its equator. The heating produces buoyancy and rotation generates a Coriolis forces in the fluid. However, due to the curved surface of the bubble, the buoyancy and Coriolis forces vary with latitude on the bubble, giving rise to rich flow behavior. We first explore the single-po… ▽ More We use Direct Numerical Simulations to study the two-dimensional flow of a rotating, half soap bubble that is heated at its equator. The heating produces buoyancy and rotation generates a Coriolis forces in the fluid. However, due to the curved surface of the bubble, the buoyancy and Coriolis forces vary with latitude on the bubble, giving rise to rich flow behavior. We first explore the single-point properties of the flow, including the Reynolds and Nusselt numbers, mean fields, and Reynolds stresses, all as a function of latitude. For a given Rayleigh number, we observe a non-monotonic dependence on the Rossby number Ro, and large scale mean circulations that are strongly influenced by rotation. We then consider quantities that reveal the multiscale nature of the flow, including spectrums and spectral fluxes of kinetic and thermal energy, and enstrophy, and structure functions of velocity and temperature. The fluxes show that just a for nonbuoyant two-dimensional turbulence on a flat surface, there is an upscale flux of kinetic energy at larger scales (fed by buoyancy injection of turbulent kinetic energy at smaller scales), and a downscale flux of enstrophy at smaller scales. The kinetic energy spectrum and velocity structure functions are well described by Bolgiano-Obukhov (BO) scaling at scales where the effects of rotation are weak. The temperature structure functions do not, however, satisfy BO scaling in general, due to strong intermittency in the temperature field. △ Less

Submitted 18 November, 2020; originally announced November 2020.

High voltage transition metal-free cathode material LiBC3F4 for Li ion batteries

Authors: Zhiqiang Wang, Lixin Xiong, Wenwei Luo, Bo Xu, Chuying Ouyang

Abstract: The structural stability and electrochemical performance of boron substituted fluorinated graphite as a Li ion batteries cathode material are studied by first principles calculations. The results show that boron substituted fluorinated graphite BC3F4 possesses excellent structural stability, good electrical and ionic conductivities. Unexpectedly, the average Li intercalation voltage of LiBC3F4 is… ▽ More The structural stability and electrochemical performance of boron substituted fluorinated graphite as a Li ion batteries cathode material are studied by first principles calculations. The results show that boron substituted fluorinated graphite BC3F4 possesses excellent structural stability, good electrical and ionic conductivities. Unexpectedly, the average Li intercalation voltage of LiBC3F4 is up to 4.44 V, which is much larger than that of LiBCF2. The average voltage of LiBC3F4 is even larger than that of common commercial transition metal oxides cathodes, indicating that LiBC3F4 is a breakthrough of transition metal-free high voltage cathode materials. By comparing the Fermi level, we found the Fermi level of LiBC3F4 is 1.38 eV lower than that of LiBCF2, leading to the decrease of the electron filling energy for the Li intercalation and forming much higher voltage. Moreover, LiBC3F4 shows small volume expansion and high energy density. LiBC3F4 is a promising high voltage cathode material for Li ion batteries. Finally, by calculating the evolution of Li intercalation voltage and Fermi level during the discharging process, a linear correlation between the Fermi level and Li intercalation voltage has been found. △ Less

Submitted 16 November, 2020; originally announced November 2020.

A variable ionized disk wind in the black-hole candidate EXO 1846-031

Authors: Yanan Wang, Long Ji, Javier A. Garcia, Thomas Dauser, Mariano Mendez, Junjie Mao, L. Tao, Diego Altamirano, Pierre Maggi, S. N. Zhang, M. Y. Ge, L. Zhang, J. L. Qu, S. Zhang, X. Ma, F. J. Lu, T. P. Li, Y. Huang, S. J. Zheng, Z. Chang, Y. L. Tuo, L. M. Song, Y. P. Xu, Y. Chen, C. Z. Liu , et al. (66 additional authors not shown)

Abstract: After 34 years, the black-hole candidate EXO 1846-031 went into outburst again in 2019. We investigate its spectral properties in the hard intermediate and the soft states with NuSTAR and Insight-HXMT. A reflection component has been detected in the two spectral states but possibly originating from different illumination spectra: in the intermediate state, the illuminating source is attributed to… ▽ More After 34 years, the black-hole candidate EXO 1846-031 went into outburst again in 2019. We investigate its spectral properties in the hard intermediate and the soft states with NuSTAR and Insight-HXMT. A reflection component has been detected in the two spectral states but possibly originating from different illumination spectra: in the intermediate state, the illuminating source is attributed to a hard coronal component, which has been commonly observed in other X-ray binaries, whereas in the soft state the reflection is probably produced by the disk self-irradiation. Both cases support EXO 1846-031 as a low inclination system of ~40 degrees. An absorption line is clearly detected at ~7.2 keV in the hard intermediate state, corresponding to a highly ionized disk wind (log ξ > 6.1) with a velocity up to 0.06c. Meanwhile, quasi-simultaneous radio emissions have been detected before and after the X-rays, implying the co-existence of disk winds and jets in this system. Additionally, the observed wind in this source is potentially driven by magnetic forces. The absorption line disappeared in the soft state and a narrow emission line appeared at ~6.7 keV on top of the reflection component, which may be evidence for disk winds, but data with the higher spectral resolution are required to examine this. △ Less

Submitted 28 December, 2020; v1 submitted 27 October, 2020; originally announced October 2020.

Comments: Accepted for publication in ApJ

STR: Secure Computation on Additive Shares Using the Share-Transform-Reveal Strategy

Authors: Zhihua Xia, Qi Gu, Wenhao Zhou, Lizhi Xiong, Jian Weng, Neal N. Xiong

Abstract: The rapid development of cloud computing has probably benefited each of us. However, the privacy risks brought by untrustworthy cloud servers arise the attention of more and more people and legislatures. In the last two decades, plenty of works seek to outsource various specific tasks while ensuring the security of private data. The tasks to be outsourced are countless; however, the computations i… ▽ More The rapid development of cloud computing has probably benefited each of us. However, the privacy risks brought by untrustworthy cloud servers arise the attention of more and more people and legislatures. In the last two decades, plenty of works seek to outsource various specific tasks while ensuring the security of private data. The tasks to be outsourced are countless; however, the computations involved are similar. In this paper, we construct a series of novel protocols that support the secure computation of various functions on numbers (e.g., the basic elementary functions) and matrices (e.g., the calculation of eigenvectors and eigenvalues) in arbitrary $n\geq 2$ servers. All protocols only require constant rounds of interactions and achieve the low computation complexity. Moreover, the proposed $n$-party protocols ensure the security of private data even though $n-1$ servers collude. The convolutional neural network models are utilized as the case studies to verify the protocols. The theoretical analysis and experimental results demonstrate the correctness, efficiency, and security of the proposed protocols. △ Less

Submitted 24 April, 2021; v1 submitted 28 September, 2020; originally announced September 2020.

Comments: This paper has been accepted by IEEE Transactions on computers

Privacy-Preserving Image Retrieval Based on Additive Secret Sharing

Authors: Zhihua Xia, Qi Gu, Lizhi Xiong, Wenhao Zhou, Jian Weng

Abstract: The rapid growth of digital images motivates individuals and organizations to upload their images to the cloud server. To preserve privacy, image owners would prefer to encrypt the images before uploading, but it would strongly limit the efficient usage of images. Plenty of existing schemes on privacy-preserving Content-Based Image Retrieval (PPCBIR) try to seek the balance between security and re… ▽ More The rapid growth of digital images motivates individuals and organizations to upload their images to the cloud server. To preserve privacy, image owners would prefer to encrypt the images before uploading, but it would strongly limit the efficient usage of images. Plenty of existing schemes on privacy-preserving Content-Based Image Retrieval (PPCBIR) try to seek the balance between security and retrieval ability. However, compared to the advanced technologies in CBIR like Convolutional Neural Network (CNN), the existing PPCBIR schemes are far deficient in both accuracy and efficiency. With more cloud service providers, the collaborative secure image retrieval service provided by multiple cloud servers becomes possible. In this paper, inspired by additive secret sharing technology, we propose a series of additive secure computing protocols on numbers and matrices with better efficiency, and then show their application in PPCBIR. Specifically, we extract CNN features, decrease the dimension of features and build the index securely with the help of our protocols, which include the full process of image retrieval in the plaintext domain. The experiments and security analysis demonstrate the efficiency, accuracy, and security of our scheme. △ Less

Submitted 18 December, 2020; v1 submitted 15 September, 2020; originally announced September 2020.

Comments: This version repairs many mistakes

Efficient Privacy-Preserving Computation Based on Additive Secret Sharing

Authors: Lizhi Xiong, Wenhao Zhou, Zhihua Xia, Qi Gu, Jian Weng

Abstract: The emergence of cloud computing provides a new computing paradigm for users -- massive and complex computing tasks can be outsourced to cloud servers. However, the privacy issues also follow. Fully homomorphic encryption shows great potential in privacy-preserving computation, yet it is not ready for practice. At present, secure multiparty computation (MPC) remains mainly approach to deal with se… ▽ More The emergence of cloud computing provides a new computing paradigm for users -- massive and complex computing tasks can be outsourced to cloud servers. However, the privacy issues also follow. Fully homomorphic encryption shows great potential in privacy-preserving computation, yet it is not ready for practice. At present, secure multiparty computation (MPC) remains mainly approach to deal with sensitive data. In this paper, following the secret sharing based MPC paradigm, we propose a secure 2-party computation scheme, in which cloud servers can securely evaluate functions with high efficiency. We first propose the multiplicative secret sharing (MSS) based on typical additive secret sharing (ASS). Then, we design protocols to switch shared secret between MSS and ASS, based on which a series of protocols for comparison and nearly all of the elementary functions are proposed. We prove that all the proposed protocols are Universally Composable secure in the honest-but-curious model. Finally, we will show the remarkable progress of our protocols on both communication efficiency and functionality completeness. △ Less

Submitted 31 March, 2021; v1 submitted 11 September, 2020; originally announced September 2020.

Comments: 12 pages, 3 figures

Constraining the transient high-energy activity of FRB180916.J0158+65 with Insight-HXMT followup observations

Authors: C. Guidorzi, M. Orlandini, F. Frontera, L. Nicastro, S. L. Xiong, J. Y. Liao, G. Li, S. N. Zhang, L. Amati, E. Virgilli, S. Zhang, Q. C. Bu, C. Cai, X. L. Cao, Z. Chang, L. Chen, T. X. Chen, Y. Chen, Y. P. Chen, W. W. Cui, Y. Y. Du, G. H. Gao, H. Gao, M. Gao, M. Y. Ge , et al. (74 additional authors not shown)

Abstract: A link between magnetars and fast radio burst (FRB) sources has finally been established. In this context, one of the open issues is whether/which sources of extra galactic FRBs exhibit X/gamma-ray outbursts and whether it is correlated with radio activity. We aim to constrain possible X/gamma-ray burst activity from one of the nearest extragalactic FRB sources currently known over a broad energy… ▽ More A link between magnetars and fast radio burst (FRB) sources has finally been established. In this context, one of the open issues is whether/which sources of extra galactic FRBs exhibit X/gamma-ray outbursts and whether it is correlated with radio activity. We aim to constrain possible X/gamma-ray burst activity from one of the nearest extragalactic FRB sources currently known over a broad energy range, by looking for bursts over a range of timescales and energies that are compatible with being powerful flares from extragalactic magnetars. We followed up the as-yet nearest extragalactic FRB source at a mere 149 Mpc distance, the periodic repeater FRB180916.J0158+65, during the active phase on February 4-7, 2020, with the Insight-Hard X-ray Modulation Telescope (HXMT). Taking advantage of the combination of broad band, large effective area, and several independent detectors available, we searched for bursts over a set of timescales from 1 ms to 1.024 s with a sensitive algorithm, that had previously been characterised and optimised. Moreover, through simulations we studied the sensitivity of our technique in the released energy-duration phase space for a set of synthetic flares and assuming different energy spectra. We constrain the possible occurrence of flares in the 1-100 keV energy band to E<10^46 erg for durations <0.1 s over several tens of ks exposure. We can rule out the occurrence of giant flares similar to the ones that were observed in the few cases of Galactic magnetars. The absence of reported radio activity during our observations does not allow us to make any statements on the possible simultaneous high-energy emission. △ Less

Submitted 27 August, 2020; originally announced August 2020.

Comments: 10 pages, 3 figures, accepted by A&A

Journal ref: A&A 642, A160 (2020)

Time-based Sequence Model for Personalization and Recommendation Systems

Authors: Tigran Ishkhanov, Maxim Naumov, Xianjie Chen, Yan Zhu, Yuan Zhong, Alisson Gusatti Azzolini, Chonglin Sun, Frank Jiang, Andrey Malevich, Liang Xiong

Abstract: In this paper we develop a novel recommendation model that explicitly incorporates time information. The model relies on an embedding layer and TSL attention-like mechanism with inner products in different vector spaces, that can be thought of as a modification of multi-headed attention. This mechanism allows the model to efficiently treat sequences of user behavior of different length. We study t… ▽ More In this paper we develop a novel recommendation model that explicitly incorporates time information. The model relies on an embedding layer and TSL attention-like mechanism with inner products in different vector spaces, that can be thought of as a modification of multi-headed attention. This mechanism allows the model to efficiently treat sequences of user behavior of different length. We study the properties of our state-of-the-art model on statistically designed data set. Also, we show that it outperforms more complex models with longer sequence length on the Taobao User Behavior dataset. △ Less

Submitted 27 August, 2020; originally announced August 2020.

Comments: 17 pages, 7 figures

MSC Class: 68T05 ACM Class: I.2.6; I.5.0; H.3.3; H.3.4

Insight-HXMT firm detection of the highest energy fundamental cyclotron resonance scattering feature in the spectrum of GRO J1008-57

Authors: M. Y. Ge, L. Ji, S. N. Zhang, A. Santangelo, C. Z. Liu, V. Doroshenko, R. Staubert, J. L. Qu, S. Zhang, F. J. Lu, L. M. Song, T. P. Li, L. Tao, Y. P. Xu, X. L. Cao, Y. Chen, Q. C. Bu, C. Cai, Z. Chang, G. Chen, L. Chen, T. X. Chen, Y. B. Chen, Y. P. Chen, W. Cui , et al. (99 additional authors not shown)

Abstract: We report on the observation of the accreting pulsar GRO J1008-57 performed by Insight-HXMT at the peak of the source's 2017 outburst. Pulsations are detected with a spin period of 93.283(1) s. The pulse profile shows double peaks at soft X-rays, and only one peak above 20 keV. The spectrum is well described by the phenomenological models of X-ray pulsars. A cyclotron resonant scattering feature i… ▽ More We report on the observation of the accreting pulsar GRO J1008-57 performed by Insight-HXMT at the peak of the source's 2017 outburst. Pulsations are detected with a spin period of 93.283(1) s. The pulse profile shows double peaks at soft X-rays, and only one peak above 20 keV. The spectrum is well described by the phenomenological models of X-ray pulsars. A cyclotron resonant scattering feature is detected with very high statistical significance at a centroid energy of $E_{\rm cyc}=90.32_{-0.28}^{+0.32}$ keV, for the reference continuum and line models, HIGHECUT and GABS respectively. Detection is very robust with respect to different continuum models. The line energy is significantly higher than what suggested from previous observations, which provided very marginal evidence for the line. This establishes a new record for the centroid energy of a fundamental cyclotron resonant scattering feature observed in accreting pulsars. We also discuss the accretion regime of the source during the Insight-HXMT observation. △ Less

Submitted 4 August, 2020; originally announced August 2020.

Comments: 8 pages, 3 figures, accepted for publication in ApJL

Correlation filter tracking with adaptive proposal selection for accurate scale estimation

Authors: Luo Xiong, Yanjie Liang, Yan Yan, Hanzi Wang

Abstract: Recently, some correlation filter based trackers with detection proposals have achieved state-of-the-art tracking results. However, a large number of redundant proposals given by the proposal generator may degrade the performance and speed of these trackers. In this paper, we propose an adaptive proposal selection algorithm which can generate a small number of high-quality proposals to handle the… ▽ More Recently, some correlation filter based trackers with detection proposals have achieved state-of-the-art tracking results. However, a large number of redundant proposals given by the proposal generator may degrade the performance and speed of these trackers. In this paper, we propose an adaptive proposal selection algorithm which can generate a small number of high-quality proposals to handle the problem of scale variations for visual object tracking. Specifically, we firstly utilize the color histograms in the HSV color space to represent the instances (i.e., the initial target in the first frame and the predicted target in the previous frame) and proposals. Then, an adaptive strategy based on the color similarity is formulated to select high-quality proposals. We further integrate the proposed adaptive proposal selection algorithm with coarse-to-fine deep features to validate the generalization and efficiency of the proposed tracker. Experiments on two benchmark datasets demonstrate that the proposed algorithm performs favorably against several state-of-the-art trackers. △ Less

Submitted 14 July, 2020; originally announced July 2020.

Comments: 6 pages, 14 figures

Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval

Authors: Lee Xiong, Chenyan Xiong, Ye Li, Kwok-Fung Tang, Jialin Liu, Paul Bennett, Junaid Ahmed, Arnold Overwijk

Abstract: Conducting text retrieval in a dense learned representation space has many intriguing advantages over sparse retrieval. Yet the effectiveness of dense retrieval (DR) often requires combination with sparse retrieval. In this paper, we identify that the main bottleneck is in the training mechanisms, where the negative instances used in training are not representative of the irrelevant documents in t… ▽ More Conducting text retrieval in a dense learned representation space has many intriguing advantages over sparse retrieval. Yet the effectiveness of dense retrieval (DR) often requires combination with sparse retrieval. In this paper, we identify that the main bottleneck is in the training mechanisms, where the negative instances used in training are not representative of the irrelevant documents in testing. This paper presents Approximate nearest neighbor Negative Contrastive Estimation (ANCE), a training mechanism that constructs negatives from an Approximate Nearest Neighbor (ANN) index of the corpus, which is parallelly updated with the learning process to select more realistic negative training instances. This fundamentally resolves the discrepancy between the data distribution used in the training and testing of DR. In our experiments, ANCE boosts the BERT-Siamese DR model to outperform all competitive dense and sparse retrieval baselines. It nearly matches the accuracy of sparse-retrieval-and-BERT-reranking using dot-product in the ANCE-learned representation space and provides almost 100x speed-up. △ Less

Submitted 20 October, 2020; v1 submitted 1 July, 2020; originally announced July 2020.

Spatio-Temporal Tensor Sketching via Adaptive Sampling

Authors: **g Ma, Qiuchen Zhang, Joyce C. Ho, Li Xiong

Abstract: Mining massive spatio-temporal data can help a variety of real-world applications such as city capacity planning, event management, and social network analysis. The tensor representation can be used to capture the correlation between space and time and simultaneously exploit the latent structure of the spatial and temporal patterns in an unsupervised fashion. However, the increasing volume of spat… ▽ More Mining massive spatio-temporal data can help a variety of real-world applications such as city capacity planning, event management, and social network analysis. The tensor representation can be used to capture the correlation between space and time and simultaneously exploit the latent structure of the spatial and temporal patterns in an unsupervised fashion. However, the increasing volume of spatio-temporal data has made it prohibitively expensive to store and analyze using tensor factorization. In this paper, we propose SkeTenSmooth, a novel tensor factorization framework that uses adaptive sampling to compress the tensor in a temporally streaming fashion and preserves the underlying global structure. SkeTenSmooth adaptively samples incoming tensor slices according to the detected data dynamics. Thus, the sketches are more representative and informative of the tensor dynamic patterns. In addition, we propose a robust tensor factorization method that can deal with the sketched tensor and recover the original patterns. Experiments on the New York City Yellow Taxi data show that SkeTenSmooth greatly reduces the memory cost and outperforms random sampling and fixed rate sampling method in terms of retaining the underlying patterns. △ Less

Submitted 21 June, 2020; originally announced June 2020.

No pulsed radio emission during a bursting phase of a Galactic magnetar

Authors: L. Lin, C. F. Zhang, P. Wang, H. Gao, X. Guan, J. L. Han, J. C. Jiang, P. Jiang, K. J. Lee, D. Li, Y. P. Men, C. C. Miao, C. H. Niu, J. R. Niu, C. Sun, B. J. Wang, Z. L. Wang, H. Xu, J. L. Xu, J. W. Xu, Y. H. Yang, Y. P. Yang, W. Yu, B. Zhang, B. -B. Zhang , et al. (23 additional authors not shown)

Abstract: Fast radio bursts (FRBs) are mysterious millisecond-duration radio transients of unknown origin observed at extragalactic distances. It has been long speculated that magnetars are the engine powering repeating bursts from FRB sources, but no convincing evidence has been collected so far\cite{sun19}. Recently, the Galactic magnetar SGR J1935+2154 entered an active phase by emitting intense soft Gam… ▽ More Fast radio bursts (FRBs) are mysterious millisecond-duration radio transients of unknown origin observed at extragalactic distances. It has been long speculated that magnetars are the engine powering repeating bursts from FRB sources, but no convincing evidence has been collected so far\cite{sun19}. Recently, the Galactic magnetar SGR J1935+2154 entered an active phase by emitting intense soft Gamma-ray bursts. One FRB-like event with two peaks (FRB 200428) and a luminosity slightly lower than the faintest extragalactic FRBs was detected from the source, in association with a soft Gamma-ray / hard X-ray flare. Here we report an eight-hour targeted radio observational campaign comprising four sessions and assisted by multi-wavelength (optical and hard X-rays) data. During the third session, 29 soft Gamma-ray repeater (SGR) bursts were detected in Gamma-ray energies. Throughout the observing period, we detected no single dispersed pulsed emission coincident with the arrivals of SGR bursts, but unfortunately we were not observing when the FRB was detected. The non-detection places a fluence upper limit that is eight orders of magnitude lower than the fluence of FRB 200428. Our results suggest that FRB -- SGR burst associations are rare. FRBs may be highly relativistic and geometrically beamed, or FRB-like events associated with SGR bursts may have narrow spectra and characteristic frequencies outside the observed band. It is also possible that the physical conditions required to achieve coherent radiation in SGR bursts are difficult to satisfy, and that only under extreme conditions could an FRB be associated with an SGR burst. △ Less

Submitted 5 November, 2020; v1 submitted 23 May, 2020; originally announced May 2020.

Comments: 27 pages, 7 figures, Published in Nature 2020, Vol 587

HXMT Identification of a non-thermal X-ray burst from SGR J1935+2154 and with FRB 200428

Authors: C. K. Li, L. Lin, S. L. Xiong, M. Y. Ge, X. B. Li, T. P. Li, F. J. Lu, S. N. Zhang, Y. L. Tuo, Y. Nang, B. Zhang, S. Xiao, Y. Chen, L. M. Song, Y. P. Xu, C. Z. Liu, S. M. Jia, X. L. Cao, J. L. Qu, S. Zhang, Y. D. Gu, J. Y. Liao, X. F. Zhao, Y. Tan, J. Y. Nie , et al. (96 additional authors not shown)

Abstract: Fast radio bursts (FRBs) are short pulses observed in radio band from cosmological distances. One class of models invoke soft gamma-ray repeaters (SGRs), or magnetars, as the sources of FRBs. Some radio pulses have been observed from some magnetars, however, no FRB-like events had been detected in association any magnetar burst, including one giant flare. Recently, a pair of FRB-like bursts (FRB 2… ▽ More Fast radio bursts (FRBs) are short pulses observed in radio band from cosmological distances. One class of models invoke soft gamma-ray repeaters (SGRs), or magnetars, as the sources of FRBs. Some radio pulses have been observed from some magnetars, however, no FRB-like events had been detected in association any magnetar burst, including one giant flare. Recently, a pair of FRB-like bursts (FRB 200428 hereafter) separated by milliseconds (ms) were detected from the general direction of the Galactic magnetar SGR J1935+2154. Here we report the detection of a non-thermal X-ray burst in the 1-250 keV energy band with the Insight-HXMT satellite, which we identify as emitted from SGR J1935+2154. The burst showed two hard peaks with a separation of 34 ms, broadly consistent with that of the two bursts in FRB 200428. The delay time between the double radio and X-ray peaks is about 8.57 s, fully consistent with the dispersion delay of FRB 200428. We thus identify the non-thermal X-ray burst is associated with FRB 200428 whose high energy counterpart is the two hard peaks in X-ray. Our results suggest that the non-thermal X-ray burst and FRB 200428 share the same physical origin in an explosive event from SGR J1935+2154. △ Less

Submitted 6 April, 2021; v1 submitted 22 May, 2020; originally announced May 2020.

Comments: 24 pages, 9 figures, 6 tables; initial submission to a journal on May 9th, 2020. Significant changes include updated localization and detailed spectral evolution of the X-ray burst, and better determination of the two narrow X-ray peaks corresponding to the two radio pulses. Conclusions are strengthened. Nature Astronomy online on Feb. 18, 2021

Journal ref: https://www.nature.com/articles/s41550-021-01302-6, Nature Astronomy online on Feb. 18, 2021

PGLP: Customizable and Rigorous Location Privacy through Policy Graph

Authors: Yang Cao, Yonghui Xiao, Shun Takagi, Li Xiong, Masatoshi Yoshikawa, Yilin Shen, **fei Liu, Hongxia **, Xiaofeng Xu

Abstract: Location privacy has been extensively studied in the literature. However, existing location privacy models are either not rigorous or not customizable, which limits the trade-off between privacy and utility in many real-world applications. To address this issue, we propose a new location privacy notion called PGLP, i.e., \textit{Policy Graph based Location Privacy}, providing a rich interface to r… ▽ More Location privacy has been extensively studied in the literature. However, existing location privacy models are either not rigorous or not customizable, which limits the trade-off between privacy and utility in many real-world applications. To address this issue, we propose a new location privacy notion called PGLP, i.e., \textit{Policy Graph based Location Privacy}, providing a rich interface to release private locations with customizable and rigorous privacy guarantee. First, we design the privacy metrics of PGLP by extending differential privacy. Specifically, we formalize a user's location privacy requirements using a \textit{location policy graph}, which is expressive and customizable. Second, we investigate how to satisfy an arbitrarily given location policy graph under adversarial knowledge. We find that a location policy graph may not always be viable and may suffer \textit{location exposure} when the attacker knows the user's mobility pattern. We propose efficient methods to detect location exposure and repair the policy graph with optimal utility. Third, we design a private location trace release framework that pipelines the detection of location exposure, policy graph repair, and private trajectory release with customizable and rigorous location privacy. Finally, we conduct experiments on real-world datasets to verify the effectiveness of the privacy-utility trade-off and the efficiency of the proposed algorithms. △ Less

Submitted 15 July, 2020; v1 submitted 4 May, 2020; originally announced May 2020.

Comments: accepted in the 25th European Symposium on Research in Computer Security (ESORICS) 2020

PANDA: Policy-aware Location Privacy for Epidemic Surveillance

Authors: Yang Cao, Shun Takagi, Yonghui Xiao, Li Xiong, Masatoshi Yoshikawa

Abstract: In this demonstration, we present a privacy-preserving epidemic surveillance system. Recently, many countries that suffer from coronavirus crises attempt to access citizen's location data to eliminate the outbreak. However, it raises privacy concerns and may open the doors to more invasive forms of surveillance in the name of public health. It also brings a challenge for privacy protection techniq… ▽ More In this demonstration, we present a privacy-preserving epidemic surveillance system. Recently, many countries that suffer from coronavirus crises attempt to access citizen's location data to eliminate the outbreak. However, it raises privacy concerns and may open the doors to more invasive forms of surveillance in the name of public health. It also brings a challenge for privacy protection techniques: how can we leverage people's mobile data to help combat the pandemic without scarifying our location privacy. We demonstrate that we can have the best of the two worlds by implementing policy-based location privacy for epidemic surveillance. Specifically, we formalize the privacy policy using graphs in light of differential privacy, called policy graph. Our system has three primary functions for epidemic surveillance: location monitoring, epidemic analysis, and contact tracing. We provide an interactive tool allowing the attendees to explore and examine the usability of our system: (1) the utility of location monitor and disease transmission model estimation, (2) the procedure of contact tracing in our systems, and (3) the privacy-utility trade-offs w.r.t. different policy graphs. The attendees can find that it is possible to have the full functionality of epidemic surveillance while preserving location privacy. △ Less

Submitted 6 June, 2020; v1 submitted 30 April, 2020; originally announced May 2020.

Comments: Accepted in the 46th International Conference on Very Large Data Bases (VLDB 2020) demonstration track

Insight-HXMT insight into switch of the accretion mode: the case of the X-ray pulsar 4U 1901+03

Authors: Y. L. Tuo, L. Ji, S. S. Tsygankov, T. Mihara, L. M. Song, M. Y. Ge, A. Nabizadeh, L. Tao, J. L. Qu, Y. Zhang, S. Zhang, S. N. Zhang, Q. C. Bu, L. Chen, Y. P. Xu, X. L. Cao, Y. Chen, C. Z. Liu, C. Cai, Z. Chang, G. Chen, T. X. Chen, Y. B. Chen, Y. P. Chen, W. Cui , et al. (98 additional authors not shown)

Abstract: We use the In data collected during the 2019 outburst from X-ray pulsar 4U 1901+03 to complement the orbital parameters reported by Fermi/GBM. Using the Insight-HXMT, we examine the correlation between the derivative of the intrinsic spin frequency and bolometric flux based on accretion torque models. It was found that the pulse profiles significantly evolve during the outburst. The existence of t… ▽ More We use the In data collected during the 2019 outburst from X-ray pulsar 4U 1901+03 to complement the orbital parameters reported by Fermi/GBM. Using the Insight-HXMT, we examine the correlation between the derivative of the intrinsic spin frequency and bolometric flux based on accretion torque models. It was found that the pulse profiles significantly evolve during the outburst. The existence of two types of the profile's pattern discovered in the Insight-HXMT data indicates that this source experienced transition between a super-critical and a sub-critical accretion regime during its 2019 outburst. Based on the evolution of the pulse profiles and the torque model, we derive the distance to 4U 1901+03 as 12.4+-0.2 kpc. △ Less

Submitted 28 April, 2020; originally announced April 2020.

Comments: 8 pages, 5 figures, accepted by JHEAP

The evolution of the broadband temporal features observed in the black-hole transient MAXI J1820+070 with Insight-HXMT

Authors: Yanan Wang, Long Ji, S. N. Zhang, Mariano Méndez, J. L. Qu, Pierre Maggi, M. Y. Ge, Erlin Qiao, L. Tao, S. Zhang, Diego Altamirano, L. Zhang, X. Ma, F. J. Lu, T. P. Li, Y. Huang, S. J. Zheng, Y. P. Chen, Z. Chang, Y. L. Tuo, C. Gungor, L. M. Song, Y. P. Xu, X. L. Cao, Y. Chen , et al. (96 additional authors not shown)

Abstract: We study the evolution of the temporal properties of MAXI 1820+070 during the 2018 outburst in its hard state from MJD 58190 to 58289 with Insight-HXMT in a broad energy band 1-150 keV. We find different behaviors of the hardness ratio, the fractional rms and time lag before and after MJD 58257, suggesting a transition occurred at around this point. The observed time lags between the soft photons… ▽ More We study the evolution of the temporal properties of MAXI 1820+070 during the 2018 outburst in its hard state from MJD 58190 to 58289 with Insight-HXMT in a broad energy band 1-150 keV. We find different behaviors of the hardness ratio, the fractional rms and time lag before and after MJD 58257, suggesting a transition occurred at around this point. The observed time lags between the soft photons in the 1-5 keV band and the hard photons in higher energy bands, up to 150 keV, are frequency-dependent: the time lags in the low-frequency range, 2-10 mHz, are both soft and hard lags with a timescale of dozens of seconds but without a clear trend along the outburst; the time lags in the high-frequency range, 1-10 Hz, are only hard lags with a timescale of tens of milliseconds; first increase until around MJD 58257 and decrease after this date. The high-frequency time lags are significantly correlated to the photon index derived from the fit to the quasi-simultaneous NICER spectrum in the 1-10 keV band. This result is qualitatively consistent with a model in which the high-frequency time lags are produced by Comptonization in a jet. △ Less

Submitted 27 April, 2020; originally announced April 2020.

Comments: Accepted for publication in ApJ

Orthogonal Over-Parameterized Training

Authors: Weiyang Liu, Rongmei Lin, Zhen Liu, James M. Rehg, Liam Paull, Li Xiong, Le Song, Adrian Weller

Abstract: The inductive bias of a neural network is largely determined by the architecture and the training algorithm. To achieve good generalization, how to effectively train a neural network is of great importance. We propose a novel orthogonal over-parameterized training (OPT) framework that can provably minimize the hyperspherical energy which characterizes the diversity of neurons on a hypersphere. By… ▽ More The inductive bias of a neural network is largely determined by the architecture and the training algorithm. To achieve good generalization, how to effectively train a neural network is of great importance. We propose a novel orthogonal over-parameterized training (OPT) framework that can provably minimize the hyperspherical energy which characterizes the diversity of neurons on a hypersphere. By maintaining the minimum hyperspherical energy during training, OPT can greatly improve the empirical generalization. Specifically, OPT fixes the randomly initialized weights of the neurons and learns an orthogonal transformation that applies to these neurons. We consider multiple ways to learn such an orthogonal transformation, including unrolling orthogonalization algorithms, applying orthogonal parameterization, and designing orthogonality-preserving gradient descent. For better scalability, we propose the stochastic OPT which performs orthogonal transformation stochastically for partial dimensions of neurons. Interestingly, OPT reveals that learning a proper coordinate system for neurons is crucial to generalization. We provide some insights on why OPT yields better generalization. Extensive experiments validate the superiority of OPT over the standard training. △ Less

Submitted 4 June, 2021; v1 submitted 9 April, 2020; originally announced April 2020.

Comments: CVPR 2021 Oral (43 Pages, Substantial Update from v3, Typos Fixed from v5)

Discovery of delayed spin-up behavior following two large glitches in the Crab pulsar, and the statistics of such processes

Authors: M. Y. Ge, S. N. Zhang, F. J. Lu, T. P. Li, J. P. Yuan, X. P. Zheng, Y. Huang, S. J. Zheng, Y. P. Chen, Z. Chang, Y. L. Tuo, Q. Cheng, C. Güngör, L. M. Song, Y. P. Xu, X. L. Cao, Y. Chen, C. Z. Liu, S. Zhang, J. L. Qu, Q. C. Bu, C. Cai, G. Chen, L. Chen, M. Z. Chen , et al. (111 additional authors not shown)

Abstract: Glitches correspond to sudden jumps of rotation frequency ($ν$) and its derivative ($\dotν$) of pulsars, the origin of which remains not well understood yet, partly because the jump processes of most glitches are not well time-resolved. There are three large glitches of the Crab pulsar, detected in 1989, 1996 and 2017, which were found to have delayed spin-up processes before the normal recovery p… ▽ More Glitches correspond to sudden jumps of rotation frequency ($ν$) and its derivative ($\dotν$) of pulsars, the origin of which remains not well understood yet, partly because the jump processes of most glitches are not well time-resolved. There are three large glitches of the Crab pulsar, detected in 1989, 1996 and 2017, which were found to have delayed spin-up processes before the normal recovery processes. Here we report two additional glitches of the Crab pulsar occurred in 2004 and 2011 for which we discovered delayed spin up processes, and present refined parameters of the largest glitch occurred in 2017. The initial rising time of the glitch is determined as $<0.48$ hour. We also carried out a statistical study of these five glitches with observed spin-up processes. The two glitches occurred in 2004 and 2011 have delayed spin-up time scales ($τ_{1}$) of $1.7\pm0.8$\,days and $1.6\pm0.4$\,days, respectively. We find that the $Δν$ vs. $|Δ{\dotν}|$ relation of these five glitches is similar to those with no detected delayed spin-up process, indicating that they are similar to the others in nature except that they have larger amplitudes. For these five glitches, the amplitudes of the delayed spin-up process ($|Δν_{\rm d1}|$) and recovery process ($Δν_{\rm d2}$), their time scales ($τ_{1}$, $τ_{2}$), and permanent changes in spin frequency ($Δν_{\rm p}$) and total frequency step ($Δν_{\rm g}$) have positive correlations. From these correlations, we suggest that the delayed spin-up processes are common for all glitches, but are too short and thus difficult to be detected for most glitches. △ Less

Submitted 1 April, 2020; originally announced April 2020.

Comments: 25 pages, 8 figures

A search for prompt gamma-ray counterparts to fast radio bursts in the Insight-HXMT data

Authors: C. Guidorzi, M. Marongiu, R. Martone, L. Nicastro, S. L. Xiong, J. Y. Liao, G. Li, S. N. Zhang, L. Amati, F. Frontera, M. Orlandini, P. Rosati, E. Virgilli, S. Zhang, Q. C. Bu, C. Cai, X. L. Cao, Z. Chang, G. Chen, L. Chen, T. X. Chen, Y. B. Chen, Y. P. Chen, W. Cui, W. W. Cui , et al. (98 additional authors not shown)

Abstract: No robust detection of prompt electromagnetic counterparts to fast radio bursts (FRBs) has yet been obtained, in spite of several multi-wavelength searches carried out so far. Specifically, X/gamma-ray counterparts are predicted by some models. We planned on searching for prompt gamma-ray counterparts in the Insight-Hard X-ray Modulation Telescope (Insight-HXMT) data, taking advantage of the uniqu… ▽ More No robust detection of prompt electromagnetic counterparts to fast radio bursts (FRBs) has yet been obtained, in spite of several multi-wavelength searches carried out so far. Specifically, X/gamma-ray counterparts are predicted by some models. We planned on searching for prompt gamma-ray counterparts in the Insight-Hard X-ray Modulation Telescope (Insight-HXMT) data, taking advantage of the unique combination of large effective area in the keV-MeV energy range and of sub-ms time resolution. We selected 39 FRBs that were promptly visible from the High-Energy (HE) instrument aboard Insight-HXMT. After calculating the expected arrival times at the location of the spacecraft, we searched for a significant excess in both individual and cumulative time profiles over a wide range of time resolutions, from several seconds down to sub-ms scales. Using the dispersion measures in excess of the Galactic terms, we estimated the upper limits on the redshifts. No convincing signal was found and for each FRB we constrained the gamma-ray isotropic-equivalent luminosity and the released energy as a function of emission timescale. For the nearest FRB source, the periodic repeater FRB180916.J0158+65, we find $L_{γ,iso}<5.5\times 10^{47}$ erg/s over 1 s, whereas $L_{γ,iso}<10^{49}-10^{51}$ erg/s for the bulk of FRBs. The same values scale up by a factor of ~100 for a ms-long emission. Even on a timescale comparable with that of the radio pulse itself no keV-MeV emission is observed. A systematic association with either long or short GRBs is ruled out with high confidence, except for subluminous events, as is the case for core-collapse of massive stars (long) or binary neutron star mergers (short) viewed off axis. Only giant flares from extra-galactic magnetars at least ten times more energetic than Galactic siblings are ruled out for the nearest FRB. △ Less

Submitted 24 March, 2020; originally announced March 2020.

Comments: 15 pages, 3 figures, 6 tables, accepted by A&A

Journal ref: A&A 637, A69 (2020)

ShadowSync: Performing Synchronization in the Background for Highly Scalable Distributed Training

Authors: Qinqing Zheng, Bor-Yiing Su, Jiyan Yang, Alisson Azzolini, Qiang Wu, Ou **, Shri Karandikar, Hagay Lupesko, Liang Xiong, Eric Zhou

Abstract: Recommendation systems are often trained with a tremendous amount of data, and distributed training is the workhorse to shorten the training time. While the training throughput can be increased by simply adding more workers, it is also increasingly challenging to preserve the model quality. In this paper, we present \shadowsync, a distributed framework specifically tailored to modern scale recomme… ▽ More Recommendation systems are often trained with a tremendous amount of data, and distributed training is the workhorse to shorten the training time. While the training throughput can be increased by simply adding more workers, it is also increasingly challenging to preserve the model quality. In this paper, we present \shadowsync, a distributed framework specifically tailored to modern scale recommendation system training. In contrast to previous works where synchronization happens as part of the training process, \shadowsync separates the synchronization from training and runs it in the background. Such isolation significantly reduces the synchronization overhead and increases the synchronization frequency, so that we are able to obtain both high throughput and excellent model quality when training at scale. The superiority of our procedure is confirmed by experiments on training deep neural networks for click-through-rate prediction tasks. Our framework is capable to express data parallelism and/or model parallelism, generic to host various types of synchronization algorithms, and readily applicable to large scale problems in other areas. △ Less

Submitted 23 February, 2021; v1 submitted 6 March, 2020; originally announced March 2020.

Occlusion Aware Unsupervised Learning of Optical Flow From Video

Authors: Jianfeng Li, Junqiao Zhao, Tiantian Feng, Chen Ye, Lu Xiong

Abstract: In this paper, we proposed an unsupervised learning method for estimating the optical flow between video frames, especially to solve the occlusion problem. Occlusion is caused by the movement of an object or the movement of the camera, defined as when certain pixels are visible in one video frame but not in adjacent frames. Due to the lack of pixel correspondence between frames in the occluded are… ▽ More In this paper, we proposed an unsupervised learning method for estimating the optical flow between video frames, especially to solve the occlusion problem. Occlusion is caused by the movement of an object or the movement of the camera, defined as when certain pixels are visible in one video frame but not in adjacent frames. Due to the lack of pixel correspondence between frames in the occluded area, incorrect photometric loss calculation can mislead the optical flow training process. In the video sequence, we found that the occlusion in the forward ($t\rightarrow t+1$) and backward ($t\rightarrow t-1$) frame pairs are usually complementary. That is, pixels that are occluded in subsequent frames are often not occluded in the previous frame and vice versa. Therefore, by using this complementarity, a new weighted loss is proposed to solve the occlusion problem. In addition, we calculate gradients in multiple directions to provide richer supervision information. Our method achieves competitive optical flow accuracy compared to the baseline and some supervised methods on KITTI 2012 and 2015 benchmarks. This source code has been released at https://github.com/jianfenglihg/UnOpticalFlow.git. △ Less

Submitted 4 March, 2020; originally announced March 2020.

Comments: 6 pages, 5 figures

ACM Class: I.4.8

Switches between accretion structures during flares in 4U 1901+03

Authors: L. Ji, L. Ducci, A. Santangelo, S. Zhang, V. Suleimanov, S. Tsygankov, V. Doroshenko, A. Nabizadeh, S. N. Zhang, M. Y. Ge, L. Tao, Q. C. Bu, J. L. Qu, F. J. Lu, L. Chen, L. M. Song, T. P. Li, Y. P. Xu, X. L. Cao, Y. Chen, C. Z. Liu, C. Cai, Z. Chang, G. Chen, T. X. Chen , et al. (98 additional authors not shown)

Abstract: We report on our analysis of the 2019 outburst of the X-ray accreting pulsar 4U 1901+03 observed with Insight-HXMT and NICER. Both spectra and pulse profiles evolve significantly in the decaying phase of the outburst. Dozens of flares are observed throughout the outburst. They are more frequent and brighter at the outburst peak. We find that the flares, which have a duration from tens to hundreds… ▽ More We report on our analysis of the 2019 outburst of the X-ray accreting pulsar 4U 1901+03 observed with Insight-HXMT and NICER. Both spectra and pulse profiles evolve significantly in the decaying phase of the outburst. Dozens of flares are observed throughout the outburst. They are more frequent and brighter at the outburst peak. We find that the flares, which have a duration from tens to hundreds of seconds, are generally brighter than the persistent emission by a factor of $\sim$ 1.5. The pulse profile shape during the flares can be significantly different than that of the persistent emission. In particular, a phase shift is clearly observed in many cases. We interpret these findings as direct evidence of changes of the pulsed beam pattern, due to transitions between the sub- and super-critical accretion regimes on a short time scale. We also observe that at comparable luminosities the flares' pulse profiles are rather similar to those of the persistent emission. This indicates that the accretion on the polar cap of the neutron star is mainly determined by the luminosity, i.e., the mass accretion rate. △ Less

Submitted 20 February, 2020; originally announced February 2020.

Comments: 11 pages, 8 figures, accepted for publication in MNRAS