arXiv subjects

Jun Zhou

Publications and source records attributed to Jun Zhou.

At least 487 records · Page 27Linked to original sources

arXiv2019

Characterizing Membership Privacy in Stochastic Gradient Langevin Dynamics

Bayesian deep learning is recently regarded as an intrinsic way to characterize the weight uncertainty of deep neural networks~(DNNs). Stochastic Gradient Langevin Dynamics~(SGLD) is an effective method to enable Bayesian deep learning on large-scale datasets. Previous theoretical studies have shown various appealing properties of SGLD, ranging from the convergence properties to the generalization bounds. In this paper, we study the properties of SGLD from a novel perspective of membership privacy protection (i.e., preventing the membership attack). The membership attack, which aims to determine whether a specific sample is used for training a given DNN model, has emerged as a common threat against deep learning algorithms. To this end, we build a theoretical framework to analyze the information leakage (w.r.t. the training dataset) of a model trained using SGLD. Based on this framework, we demonstrate that SGLD can prevent the information leakage of the training dataset to a certain extent. Moreover, our theoretical analysis can be naturally extended to other types of Stochastic Gradient Markov Chain Monte Carlo (SG-MCMC) methods. Empirical results on different datasets and models verify our theoretical findings and suggest that the SGLD algorithm can not only reduce the information leakage but also improve the generalization ability of the DNN models in real-world applications.

cs.LG↗

arXiv2019

Pruning from Scratch

Network pruning is an important research field aiming at reducing computational costs of neural networks. Conventional approaches follow a fixed paradigm which first trains a large and redundant network, and then determines which units (e.g., channels) are less important and thus can be removed. In this work, we find that pre-training an over-parameterized model is not necessary for obtaining the target pruned structure. In fact, a fully-trained over-parameterized model will reduce the search space for the pruned structure. We empirically show that more diverse pruned structures can be directly pruned from randomly initialized weights, including potential models with better performance. Therefore, we propose a novel network pruning pipeline which allows pruning from scratch. In the experiments for compressing classification models on CIFAR10 and ImageNet datasets, our approach not only greatly reduces the pre-training burden of traditional pruning methods, but also achieves similar or even higher accuracy under the same computation budgets. Our results facilitate the community to rethink the effectiveness of existing techniques used for network pruning.

cs.CV↗

arXiv2019

Generalization in Generative Adversarial Networks: A Novel Perspective from Privacy Protection

In this paper, we aim to understand the generalization properties of generative adversarial networks (GANs) from a new perspective of privacy protection. Theoretically, we prove that a differentially private learning algorithm used for training the GAN does not overfit to a certain degree, i.e., the generalization gap can be bounded. Moreover, some recent works, such as the Bayesian GAN, can be re-interpreted based on our theoretical insight from privacy protection. Quantitatively, to evaluate the information leakage of well-trained GAN models, we perform various membership attacks on these models. The results show that previous Lipschitz regularization techniques are effective in not only reducing the generalization gap but also alleviating the information leakage of the training dataset.

cs.LG↗

arXiv2019

Information collection for fraud detection in P2P financial market

Fintech companies have been facing challenges from fraudulent behavior for a long time. Fraud rate in Chinese P2P financial market could go as high as 10%. It is crucial to collect sufficient information of the user as input to the anti-fraud process. Data collection framework for Fintech companies are different fro m conventional internet firms. With individual-based crawling request , we need to deal with new challenges negligible elsewhere . In this paper , we give an outline of how we collect data from the web to facilitate our anti-fraud process. We also overview the challenges and solutions to our problems. Our team at HC Financial Service Group is one of the few companies that are capable of developing full-fledged crawlers on our own.

cs.OH↗

arXiv2019

A Thermal Resistance Network Model for Heat Conduction of Amorphous Polymers

Thermal conductivities (TCs) of the vast majority of amorphous polymers are in a very narrow range, 0.1 $\sim$ 0.5 Wm$^{-1}$K$^{-1}$, although single polymer chains possess TC of orders-of-magnitude higher. Entanglement of polymer chains plays an important role in determining the TC of bulk polymers. We propose a thermal resistance network (TRN) model for TC in amorphous polymers taking into account the entanglement of molecular chains. Our model explains well the physical origin of universally low TC observed in amorphous polymers. The empirical formulae of pressure and temperature dependence of TC can be successfully reproduced from our model not only in solid polymers but also in polymer melts. We further quantitatively explain the anisotropic TC in oriented polymers.

cond-mat.soft↗

arXiv2019

Design triple points, nexus points and related topological phases by stacking monolayers

Triple points and nexus points are two interesting topological phases, which have been reported in some three-dimensional (3D) materials. Here, we propose that triple points, nexus points and related phases, such as topological tangle nodal lines, can be obtained by alternatively stacking two types of monolayers. Two conditions for the stacking monolayers are required: the first condition is that they have a three-fold (C3) rotation symmetry and three mirror planes along the C3 axis; the second condition is that one of the monolayers should be insulating while the other one should be metallic (or semiconducting) and has a double degenerate band and a nondegenerate band at Γpoint around the Fermi level. Hexagonal boron nitride (HBN) and α/α^'-boron sheets (α/α^'-BS) are suggested as candidate materials. Even if HBN is a wide-gap insulator, the interactions between layers lead to crossings of the nondegenerate and double degenerate bands along the direction normal to the nanosheets, and thus form triple/nexus points or related phases. A tight-binding model is adopted to explain the phase transition between triple points, nexus points and other related phases.

cond-mat.mtrl-sci↗

arXiv2019

Evidence for metallic 1T phase, 3d1 electronic configuration and charge density wave order in molecular-beam epitaxy grown monolayer VTe2

We present a combined experimental and theoretical study of monolayer VTe2 grown on highly oriented pyrolytic graphite by molecular-beam epitaxy. Using various in-situ microscopic and spectroscopic techniques, including scanning tunneling microscopy/spectroscopy, synchrotron X-ray and angle-resolved photoemission, and X-ray absorption, together with theoretical analysis by density functional theory calculations, we demonstrate direct evidence of the metallic 1T phase and 3d1 electronic configuration in monolayer VTe2 that also features a (4 x 4) charge density wave order at low temperatures. In contrast to previous theoretical predictions, our element-specific characterization by X-ray magnetic circular dichroism rules out a ferromagnetic order intrinsic to the monolayer. Our findings provide essential knowledge necessary for understanding this interesting yet less explored metallic monolayer in the emerging family of van der Waals magnets.

cond-mat.mtrl-sci↗

arXiv2019

Material Based Object Tracking in Hyperspectral Videos: Benchmark and Algorithms

Traditional color images only depict color intensities in red, green and blue channels, often making object trackers fail in challenging scenarios, e.g., background clutter and rapid changes of target appearance. Alternatively, material information of targets contained in a large amount of bands of hyperspectral images (HSI) is more robust to these difficult conditions. In this paper, we conduct a comprehensive study on how material information can be utilized to boost object tracking from three aspects: benchmark dataset, material feature representation and material based tracking. In terms of benchmark, we construct a dataset of fully-annotated videos, which contain both hyperspectral and color sequences of the same scene. Material information is represented by spectral-spatial histogram of multidimensional gradient, which describes the 3D local spectral-spatial structure in an HSI, and fractional abundances of constituted material components which encode the underlying material distribution. These two types of features are embedded into correlation filters, yielding material based tracking. Experimental results on the collected benchmark dataset show the potentials and advantages of material based object tracking.

cs.CV↗

arXiv2019

TitAnt: Online Real-time Transaction Fraud Detection in Ant Financial

With the explosive growth of e-commerce and the booming of e-payment, detecting online transaction fraud in real time has become increasingly important to Fintech business. To tackle this problem, we introduce the TitAnt, a transaction fraud detection system deployed in Ant Financial, one of the largest Fintech companies in the world. The system is able to predict online real-time transaction fraud in mere milliseconds. We present the problem definition, feature extraction, detection methods, implementation and deployment of the system, as well as empirical effectiveness. Extensive experiments have been conducted on large real-world transaction data to show the effectiveness and the efficiency of the proposed system.

cs.LG↗

arXiv2019

A One-step Pruning-recovery Framework for Acceleration of Convolutional Neural Networks

Acceleration of convolutional neural network has received increasing attention during the past several years. Among various acceleration techniques, filter pruning has its inherent merit by effectively reducing the number of convolution filters. However, most filter pruning methods resort to tedious and time-consuming layer-by-layer pruning-recovery strategy to avoid a significant drop of accuracy. In this paper, we present an efficient filter pruning framework to solve this problem. Our method accelerates the network in one-step pruning-recovery manner with a novel optimization objective function, which achieves higher accuracy with much less cost compared with existing pruning methods. Furthermore, our method allows network compression with global filter pruning. Given a global pruning rate, it can adaptively determine the pruning rate for each single convolutional layer, while these rates are often set as hyper-parameters in previous approaches. Evaluated on VGG-16 and ResNet-50 using ImageNet, our approach outperforms several state-of-the-art methods with less accuracy drop under the same and even much fewer floating-point operations (FLOPs).

cs.CV↗

arXiv2019

Ultralow lattice thermal conductivity and electronic properties of monolayer 1T phase semimetal SiTe2 and SnTe2

2H phase (trigonal prismatic D3h) of layered two-dimensional (2D) transition metal dichalcogenides (TMDs) have attracted a lot of interests due to the superior electronic and optoelectronic properties. However, flexible electronic devices and thermoelectric performances based on 2H phase have been potentially limited by the strain sensitive electronic band gap and high lattice thermal conductivity (kappa_L). Here, we predict and calculate two 1T (octahedral Oh) phase monolayer telluride materials SnTe2 and SiTe2 with soft mechanics, ultralow kappa_L and electronic properties. The calculated in-plane Young's modulus of monolayer SnTe2 is softer than most of 1T-MX2 compounds. Furthermore, monolayer SiTe2 and SnTe2 also have relatively flexible electronic properties under large biaxial strain, indicating potential flexible electrode materials. Meanwhile, monolayer SiTe2 and SnTe2 both exhibit ultralow \k{appa}_L (2.27 W/mK of SiTe2 and 1.62 W/mK of SnTe2) at room temperature. Considering both acoustic and polar optical phonon scattering of the electronic relaxation time, the figure of merit (ZT) can achieve 0.46 at 600 K and 0.71 at 900 K for monolayer SiTe2 and SnTe2 respectively.

physics.comp-ph↗

arXiv2019

Intrinsic Ferromagnetism in Electrenes

We report intrinsic ferromagnetism in monolayer electrides or electrenes, in which excess electrons act as anions. Our first-principles calculations demonstrate that magnetism in such electron-rich two-dimensional (2D) materials originates from the anionic electrons rather than partially filled d orbitals, which is fundamentally different from ferromagnetism found in other 2D intrinsic magnetic materials. Taking the honeycomb LaBr$_2$ (La$^{3+}$Br$^{-}_{2}\cdot e^{-}$) as an example, our calculations reveal that the excess electron is localized at the center of the hexagon, which leads to strong Stoner-instability of the associated states at the Fermi energy, resulting in spontaneous magnetization and formation of a local moment. The overlap of extended tails of the wave functions of these electrons mediates a long-range ferromagnetic interaction, contributing to a Curie temperature ($T_\textrm{c}$) of 235 K and a coercive field ($H_\textrm{c}$) of 0.53 T, which can be further enhanced by hole doping. The dual nature, localization and extension, of the electronic states suggests a unique mechanism in such magnetic-element-free electrenes as intrinsic 2D ferromagnets.

physics.comp-ph↗

arXiv2019

2DMatPedia: An open computational database of two-dimensional materials from top-down and bottom-up approaches

Two-dimensional (2D) materials have been a hot research topic in the last decade, due to novel fundamental physics in the reduced dimension and appealing applications. Systematic discovery of functional 2D materials has been the focus of many studies. Here, we present a large dataset of 2D materials, with more than 6,000 monolayer structures, obtained from both top-down and bottom-up discovery procedures. First, we screened all bulk materials in the database of Materials Project for layered structures by a topology-based algorithm, and theoretically exfoliate them into monolayers. Then, we generated new 2D materials by chemical substitution of elements in known 2D materials by others from the same group in the periodic table. The structural, electronic and energetic properties of these 2D materials are consistently calculated, to provide a starting point for further material screening, data mining, data analysis and artificial intelligence applications. We present the details of computational methodology, data record and technical validation of our publicly available data (http://www.2dmatpedia.org/).

cond-mat.mtrl-sci↗

arXiv2019

Discovery of Hidden Classes of Layered Electrides by Extensive High-throughput Materials Screening

Despite their extraordinary properties, electrides are still a relatively unexplored class of materials with only a few compounds grown experimentally. Especially for layered electrides, the current researches mainly focus on several isostructures of Ca2N with similar interlayer two-dimensional (2D) anionic electrons. An extensive screening for different layered electrides is still missing. Here, by screening materials with anionic electrons for the structures in Materials Project, we uncover 12 existing materials as new layered electrides. Remarkably, these layered electrides demonstrate completely different properties from Ca2N. For example, unusual fully spin-polarized zero-dimensional (0D) anionic electrons are shown in metal halides with MoS2-like structures; unique one-dimensional (1D) anionic electrons are confined within the tubes of the quasi-1D structures; a coexistence of magnetic and non-magnetic anionic electrons is found in ZrCl-like structures and a new ternary Ba2LiN with both 0D and 1D anionic electrons. These materials not only significantly increase the pool of experimentally synthesizable layered electrides but also are promising to be exfoliated into advanced 2D materials.

cond-mat.mtrl-sci↗

arXiv2018

GeniePath: Graph Neural Networks with Adaptive Receptive Paths

We present, GeniePath, a scalable approach for learning adaptive receptive fields of neural networks defined on permutation invariant graph data. In GeniePath, we propose an adaptive path layer consists of two complementary functions designed for breadth and depth exploration respectively, where the former learns the importance of different sized neighborhoods, while the latter extracts and filters signals aggregated from neighbors of different hops away. Our method works in both transductive and inductive settings, and extensive experiments compared with competitive methods show that our approaches yield state-of-the-art results on large graphs.

cs.LG↗

arXiv2018

Object Tracking in Hyperspectral Videos with Convolutional Features and Kernelized Correlation Filter

Target tracking in hyperspectral videos is a new research topic. In this paper, a novel method based on convolutional network and Kernelized Correlation Filter (KCF) framework is presented for tracking objects of interest in hyperspectral videos. We extract a set of normalized three-dimensional cubes from the target region as fixed convolution filters which contain spectral information surrounding a target. The feature maps generated by convolutional operations are combined to form a three-dimensional representation of an object, thereby providing effective encoding of local spectral-spatial information. We show that a simple two-layer convolutional networks is sufficient to learn robust representations without the need of offline training with a large dataset. In the tracking step, KCF is adopted to distinguish targets from neighboring environment. Experimental results demonstrate that the proposed method performs well on sample hyperspectral videos, and outperforms several state-of-the-art methods tested on grayscale and color videos in the same scene.

cs.CV↗

arXiv2018

Wearable Affective Robot

With the development of the artificial intelligence (AI), the AI applications have influenced and changed people's daily life greatly. Here, a wearable affective robot that integrates the affective robot, social robot, brain wearable, and wearable 2.0 is proposed for the first time. The proposed wearable affective robot is intended for a wide population, and we believe that it can improve the human health on the spirit level, meeting the fashion requirements at the same time. In this paper, the architecture and design of an innovative wearable affective robot, which is dubbed as Fitbot, are introduced in terms of hardware and algorithm's perspectives. In addition, the important functional component of the robot-brain wearable device is introduced from the aspect of the hardware design, EEG data acquisition and analysis, user behavior perception, and algorithm deployment, etc. Then, the EEG based cognition of user's behavior is realized. Through the continuous acquisition of the in-depth, in-breadth data, the Fitbot we present can gradually enrich user's life modeling and enable the wearable robot to recognize user's intention and further understand the behavioral motivation behind the user's emotion. The learning algorithm for the life modeling embedded in Fitbot can achieve better user's experience of affective social interaction. Finally, the application service scenarios and some challenging issues of a wearable affective robot are discussed.

cs.HC↗

arXiv2018

Semi-dense Stereo Matching using Dual CNNs

A robust solution for semi-dense stereo matching is presented. It utilizes two CNN models for computing stereo matching cost and performing confidence-based filtering, respectively. Compared to existing CNNs-based matching cost generation approaches, our method feeds additional global information into the network so that the learned model can better handle challenging cases, such as lighting changes and lack of textures. Through utilizing non-parametric transforms, our method is also more self-reliant than most existing semi-dense stereo approaches, which rely highly on the adjustment of parameters. The experimental results based on Middlebury Stereo dataset demonstrate that the proposed approach outperforms the state-of-the-art semi-dense stereo approaches.

cs.CV↗