arXiv subjects

Meng Zhang

Publications and source records attributed to Meng Zhang.

At least 235 records · Page 13Linked to original sources

arXiv2019

Deep-seismic-prior-based reconstruction of seismic data using convolutional neural networks

Reconstruction of seismic data with missing traces is a long-standing issue in seismic data processing. In recent years, rank reduction operations are being commonly utilized to overcome this problem, which require the rank of seismic data to be a prior. However, the rank of field data is unknown; usually it requires much time to manually adjust the rank and just obtain an approximated rank. Methods based on deep learning require very large datasets for training; however acquiring large datasets is difficult owing to physical or financial constraints in practice. Therefore, in this work, we developed a novel method based on unsupervised learning using the intrinsic properties of a convolutional neural network known as U-net, without training datasets. Only one undersampled seismic data was needed, and the deep seismic prior of input data could be exploited by the network itself, thus making the reconstruction convenient. Furthermore, this method can handle both irregular and regular seismic data. Synthetic and field data were tested to assess the performance of the proposed algorithm (DSPRecon algorithm); the advantages of using our method were evaluated by comparing it with the singular spectrum analysis (SSA) method for irregular data reconstruction and de-aliased Cadzow method for regular data reconstruction. Experimental results showed that our method provided better reconstruction performance than the SSA or Cadzow methods. The recovered signal-to-noise ratios (SNRs) were 32.68 dB and 19.11 dB for the DSPRecon and SSA algorithms, respectively. Those for the DSPRecon and Cadzow methods were 35.91 dB and 15.32 dB, respectively.

cs.LG↗

arXiv2019

Seismic data interpolation based on U-net with texture loss

Missing traces in acquired seismic data is a common occurrence during the collection of seismic data. Deep neural network (DNN) has shown considerable promise in restoring incomplete seismic data. However, several DNN-based approaches ignore the specific characteristics of seismic data itself, and only focus on reducing the difference between the recovered and the original signals. In this study, a novel Seismic U-net InterpolaTor (SUIT) is proposed to preserve the seismic texture information while reconstructing the missing traces. Aside from minimizing the reconstruction error, SUIT enhances the texture consistency between the recovery and the original completely seismic data, by designing a pre-trained U-Net to extract the texture information. The experiments show that our method outperforms the classic state-of-art methods in terms of robustness.

physics.geo-ph↗

arXiv2019

Wireless Power Transfer with Information Asymmetry: A Public Goods Perspective

Wireless power transfer (WPT) technology enables a cost-effective and sustainable energy supply in wireless networks. However, the broadcast nature of wireless signals makes them non-excludable public goods, which leads to potential free-riders among energy receivers. In this study, we formulate the wireless power provision problem as a public goods provision problem, aiming to maximize the social welfare of a system of an energy transmitter (ET) and all the energy users (EUs), while considering their private information and self-interested behaviors. We propose a two-phase all-or-none scheme involving a low-complexity Power And Taxation (PAT) mechanism, which ensures voluntary participation, truthfulness, budget balance, and social optimality at every Nash equilibrium (NE). We propose a distributed PAT (D-PAT) algorithm to reach an NE, and prove its convergence by connecting the structure of NEs and that of the optimal solution to a related optimization problem. We further extend the analysis to a multi-channel system, which brings a further challenge due to the non-strict concavity of the agents' payoffs. We propose a Multi-Channel PAT (M-PAT) mechanism and a distributed M-PAT (D-MPAT) algorithm to address the challenge. Simulation results show that our design is most beneficial when there are more EUs with more homogeneous channel gains.

cs.NI↗

arXiv2019

Efficient Network Sharing with Asymmetric Constraint Information

Network sharing has become a key feature of various enablers of the next generation network, such as network function virtualization and fog computing architectures. Network utility maximization (NUM) is a general framework for achieving fair, efficient, and cost-effective sharing of constrained network resources. When agents have asymmetric and private information, however, a fundamental economic challenge is how to solve the NUM Problem considering the self-interests of strategic agents. Many previous related works have proposed economic mechanisms that can cope with agents' private utilities. However, the network sharing paradigm introduces the issue of information asymmetries regarding constraints. The related literature largely neglected such an issue; limited closely related studies provided solutions only applicable to specific application scenarios. To tackle these issues, we propose the Decomposable NUM (DeNUM) Mechanism and the Dynamic DeNUM (DyDeNUM) Mechanism, the first mechanisms in the literature for solving NUM Problems considering private utility and constraint information. The key idea of both mechanisms is to decentralize the decision process to agents, who will make resource allocation decisions without the need of revealing private information to others. Under a monitorable influence assumption, the DeNUM Mechanism yields the network-utility maximizing solution at an equilibrium, and achieves other desirable economic properties (such as individual rationality and budget balance). We further establish the connection between the equilibrium structure and the primal-dual solution to a related optimization problem, based on which we prove the convergence of the DeNUM Algorithm to an equilibrium. When the agents' influences are not monitorable, we propose the DyDeNUM Mechanism that yields the network-utility maximizing solution at the cost of the balanced budget.

cs.NI↗

arXiv2019

A Survey of Data Fusion in Smart City Applications

The advancement of various research sectors such as Internet of Things (IoT), Machine Learning, Data Mining, Big Data, and Communication Technology has shed some light in transforming an urban city integrating the aforementioned techniques to a commonly known term - Smart City. With the emergence of smart city, plethora of data sources have been made available for wide variety of applications. The common technique for handling multiple data sources is data fusion, where it improves data output quality or extracts knowledge from the raw data. In order to cater evergrowing highly complicated applications, studies in smart city have to utilize data from various sources and evaluate their performance based on multiple aspects. To this end, we introduce a multi-perspectives classification of the data fusion to evaluate the smart city applications. Moreover, we applied the proposed multi-perspectives classification to evaluate selected applications in each domain of the smart city. We conclude the paper by discussing potential future direction and challenges of data fusion integration.

eess.SP↗

arXiv2019

How to Price Fresh Data

We introduce the concept of a fresh data market, in which a destination user requests, and pays for, fresh data updates from a source provider. Data freshness is captured by the {\it age of information} (AoI) metric, defined as the time elapsed since the latest update has reached the destination. The source incurs an operational cost, modeled as an increasing convex function of the number of updates. The destination incurs an age-related cost, modeled as an increasing convex function of the AoI. The source charges the destination for each update and designs a pricing mechanism to maximize its profit; the destination on the other hand chooses a data update schedule to minimize the summation of its payments to the source and its age-related cost. The interaction among the source and destination is hence game-theoretic. Motivated by the existing pricing literature, we first study a time-dependent pricing scheme, in which the price for each update depends on when it is requested. We show in this case that the game equilibrium leads to only one data update, which does not yield the maximum profit to the source. This motivates us to consider a quantity-based pricing scheme, in which the price of each update depends on how many updates have been previously requested. We show that among all pricing schemes in which the price of an update may vary according to both time and quantity, the quantity-based pricing scheme performs best: it maximizes the source's profit and minimizes the social cost of the system, defined as the aggregate source's operational cost and the destination's age-related cost. Numerical results show that the optimal quantity-based pricing can be 27% more profitable for the source and incurs 54% less social cost, compared with the optimal time-dependent pricing.

cs.NI↗

arXiv2019

Hybrid Pricing for Mobile Collaborative Internet Access

Mobile Collaborative Internet Access (MCA) enables mobile users to share their Internet through flexible tethering arrangements. This can potentially make better use of network resources. However, from a mobile network operator's (MNO's) viewpoint, it can either reduce revenue or increase congestion, and thus has been blocked by some MNOs in practice. We propose a hybrid pricing framework for MNOs who charge users separately for access and tethering. This scheme serves to coordinate the tethering decisions of mobile users with MNO network management objectives. We analyze the MNOs' equilibrium pricing strategies in both cooperative and competitive scenarios. In the cooperative scenario, at the equilibrium, each user's cost is independent of any chosen tethering links. We then characterize the optimal hybrid pricing strategies of MNOs in this scenario. For the competitive scenario, we formulate the MNOs' competitive interactions as a pricing game, and we show that MNO competition leads to equalized prices for users if an equilibrium exists but does not guarantee its existence. Both insights motivate a quantity competition game, which is shown to guarantee an equilibrium. Simulation results show that in scenarios of interest the proposed hybrid pricing schemes can double both MNOs' profit and users' payoff and such improvements increase with the degree of network heterogeneity.

cs.NI↗

arXiv2019

Interfacial superconductivity induced by single-quintuple-layer Bi$_2$Te$_3$ on top of FeTe forming van-der-Waals heterostructure

We report the first clear observation of interfacial superconductivity on top of FeTe(FT) covered by one quintuple-layer Bi$_2$Te$_3$(BT) forming van-der-Waals heterojunction. Both transport and scanning tunneling spectroscopy measurements confirm the occurrence of superconductivity at a transition temperature T$_c$ = 13~K, when a single-quintuple-layer BT is deposited on the non-superconducting FT surface. The superconductivity gap decays exponentially with the thickness of BT, suggesting it occurs at the BT-FT interface and the proximity length is above 5-6~nm. We also measure the work function's dependence on the thickness of BT, implying a charge transfer may occur at the BT/FT interface to introduce hole doping into the FT layer, which may serve as a possible candidate for the superconducting mechanism. Our BT/FT heterojunction provides a clean system to study the unconventional interfacial superconductivity.

cond-mat.supr-con↗

arXiv2019

Multi-Granularity Reasoning for Social Relation Recognition from Images

Discovering social relations in images can make machines better interpret the behavior of human beings. However, automatically recognizing social relations in images is a challenging task due to the significant gap between the domains of visual content and social relation. Existing studies separately process various features such as faces expressions, body appearance, and contextual objects, thus they cannot comprehensively capture the multi-granularity semantics, such as scenes, regional cues of persons, and interactions among persons and objects. To bridge the domain gap, we propose a Multi-Granularity Reasoning framework for social relation recognition from images. The global knowledge and mid-level details are learned from the whole scene and the regions of persons and objects, respectively. Most importantly, we explore the fine-granularity pose keypoints of persons to discover the interactions among persons and objects. Specifically, the pose-guided Person-Object Graph and Person-Pose Graph are proposed to model the actions from persons to object and the interactions between paired persons, respectively. Based on the graphs, social relation reasoning is performed by graph convolutional networks. Finally, the global features and reasoned knowledge are integrated as a comprehensive representation for social relation recognition. Extensive experiments on two public datasets show the effectiveness of the proposed framework.

cs.CV↗

arXiv2018

Predicting Inpatient Discharge Prioritization With Electronic Health Records

Identifying patients who will be discharged within 24 hours can improve hospital resource management and quality of care. We studied this problem using eight years of Electronic Health Records (EHR) data from Stanford Hospital. We fit models to predict 24 hour discharge across the entire inpatient population. The best performing models achieved an area under the receiver-operator characteristic curve (AUROC) of 0.85 and an AUPRC of 0.53 on a held out test set. This model was also well calibrated. Finally, we analyzed the utility of this model in a decision theoretic framework to identify regions of ROC space in which using the model increases expected utility compared to the trivial always negative or always positive classifiers.

cs.LG↗

arXiv2018

Hair-GANs: Recovering 3D Hair Structure from a Single Image

We introduce Hair-GANs, an architecture of generative adversarial networks, to recover the 3D hair structure from a single image. The goal of our networks is to build a parametric transformation from 2D hair maps to 3D hair structure. The 3D hair structure is represented as a 3D volumetric field which encodes both the occupancy and the orientation information of the hair strands. Given a single hair image, we first align it with a bust model and extract a set of 2D maps encoding the hair orientation information in 2D, along with the bust depth map to feed into our Hair-GANs. With our generator network, we compute the 3D volumetric field as the structure guidance for the final hair synthesis. The modeling results not only resemble the hair in the input image but also possesses many vivid details in other views. The efficacy of our method is demonstrated by using a variety of hairstyles and comparing with the prior art.

cs.GR↗

arXiv2018

Cooperative Communication based Connectivity Recovery for UAV Networks

UAV networks often partition into separated clusters due to the high node and link dynamic. As a result, network connectivity recovery is an important issue in this area. Existing solutions always need excessive movement of nodes and thus lead to low recovery efficiency in terms of the time and energy consumption. In this paper, we for the first time study the issue of how to utilize cooperative communication technology to improve the connectivity recovery efficiency in UAV networks. We propose a Cooperative Communication based Connectivity Recovery algorithm for UAV Networks, named C3RUN. The key novelty is C3RUN not only uses cooperative communication to enlarge node's communication range and thus achieve quick repair of network connectivity, but also enables nodes to proactively move to better places for ensuring the establishment of cooperative communication links. We conduct extensive simulations to evaluate the performance of C3RUN. The simulation results reveal that C3RUN can not only achieve connectivity recovery with less nodes and shorter distance to move, but also always finish recovery with less time, when comparing with existing work. Furthermore, C3RUN can achieve 100% success ratio for connectivity recovery.

cs.NI↗

arXiv2018

Superconductivity in a misfit layered compound (SnSe)$_{1.16}$(NbSe$_2$)

The large size single crystals of (SnSe)$_{1.16}$(NbSe$_2$) misfit layered compound were grown and superconductivity with $T_c$ of 3.4 K was first discovered in this system. Powder X-ray diffraction (XRD) and high resolution transmission electron microscopy (HRTEM) clearly display the misfit feature between SnSe and NbSe$_2$ subsystems. The Sommerfeld coefficient $γ$ inferred from specific-heat measurements is 16.73 mJ mol$^{-1}$ K$^{-2}$, slightly larger than the usual misfit compounds. The normalized specific heat jump $Δ$$C_e$/$γ$$T_{\rm c}$ is about 0.98, and the electron-phonon coupling constant $λ$$_{e-ph}$ is estimated to be 0.80. The estimated value of the in-plane upper critical magnetic field, $H_{c2}^{ab}$(0), is about 7.82 T, exceeding the Pauli paramagnetic limit slightly. Both the specific-heat and $H_{c2}$ data suggest that (SnSe)$_{1.16}$(NbSe$_2$) is a multi-band superconductor.

cond-mat.supr-con↗

arXiv2018

Recent Advances in Convolutional Neural Network Acceleration

In recent years, convolutional neural networks (CNNs) have shown great performance in various fields such as image classification, pattern recognition, and multi-media compression. Two of the feature properties, local connectivity and weight sharing, can reduce the number of parameters and increase processing speed during training and inference. However, as the dimension of data becomes higher and the CNN architecture becomes more complicated, the end-to-end approach or the combined manner of CNN is computationally intensive, which becomes limitation to CNN's further implementation. Therefore, it is necessary and urgent to implement CNN in a faster way. In this paper, we first summarize the acceleration methods that contribute to but not limited to CNN by reviewing a broad variety of research papers. We propose a taxonomy in terms of three levels, i.e.~structure level, algorithm level, and implementation level, for acceleration methods. We also analyze the acceleration methods in terms of CNN architecture compression, algorithm optimization, and hardware-based improvement. At last, we give a discussion on different perspectives of these acceleration and optimization methods within each level. The discussion shows that the methods in each level still have large exploration space. By incorporating such a wide range of disciplines, we expect to provide a comprehensive reference for researchers who are interested in CNN acceleration.

cs.LG↗

arXiv2018

Stellar mass distribution and star formation history of the Galactic disk revealed by mono-age stellar populations from LAMOST

We present a detailed determination and analysis of 3D stellar mass distribution of the Galactic disk for mono-age populations using a sample of 0.93 million main-sequence turn-off and subgiant stars from the LAMOST Galactic Surveys. Our results show (1) all stellar populations younger than 10\,Gyr exhibit strong disk flaring, which is accompanied with a dumpy vertical density profile that is best described by a $sech^n$ function with index depending on both radius and age; (2) Asymmetries and wave-like oscillations are presented in both the radial and vertical direction, with strength varying with stellar populations; (3) As a contribution by the Local spiral arm, the mid-plane stellar mass density at solar radius but 400--800\,pc (3--6$^\circ$) away from the Sun in the azimuthal direction has a value of $0.0594\pm0.0008$\,$M_\odot$/pc$^3$, which is 0.0164\,$M_\odot$/pc$^3$ higher than previous estimates at the solar neighborhood. The result causes doubts on the current estimate of local dark matter density; (4) The radial distribution of surface mass density yields a disk scale length evolving from $\sim$4\,kpc for the young to $\sim$2\,kpc for the old populations. The overall population exhibits a disk scale length of $2.48\pm0.05$\,kpc, and a total stellar mass of $3.6(\pm0.1)\times10^{10}$\,$M_\odot$ assuming $R_{\odot}=8.0$\,kpc, and the value becomes $4.1(\pm0.1)\times10^{10}$\,$M_\odot$ if $R_{\odot}=8.3$\,kpc; (5) The disk has a peak star formation rate ({\rm SFR}) changing from 6--8\,Gyr at the inner to 4--6\,Gyr ago at the outer part, indicating an inside-out assemblage history. The 0--1\,Gyr population yields a recent disk total {\rm SFR} of $1.96\pm0.12$\,$M_\odot$/yr.

astro-ph.GA↗

arXiv2018

Origin of interfacial conductivity at complex oxide heterointerfaces: possibility of electron transfer from water chemistry at surface oxygen vacancies

Variety of conducting heterointerfaces have been made between SrTiO3 substrates and thin capping layers of distinctly different oxide materials that can be classified into polar band insulators (e.g. LaAlO3), polar Mott insulators (e.g. LaTiO3), apparently non-polar band insulators (e.g. γ-Al2O3), and amorphous oxides (e.g. amorphous SrTiO3). A fundamental question to ask is if there is a common mechanism that governs interfacial conductivity in all these heterointerfaces. Here, we examined the conductivity of different kinds of heterointerfaces by annealing in oxygen and surface treatment with water. It was found that the conductivity of all the heterointerfaces show a strong dependence on annealing, and can be universally tuned by surface treatment whose effect is determined by the annealing condition. These observations, together with ambient-pressure X-ray photoelectron spectroscopy measurements, suggest that water chemistry at surface oxygen vacancies is a common mechanism that supplies electrons to the interface.

cond-mat.mtrl-sci↗

arXiv2018

Bearing fault diagnosis based on domain adaptation using transferable features under different working conditions

Bearing failure is the most common failure mode in rotating machinery and can result in large financial losses or even casualties. However, complex structures around bearing and actual variable working conditions can lead to large distribution difference of vibration signal between a training set and a test set, which causes the accuracy-dropping problem of fault diagnosis. Thus, how to improve efficiently the performance of bearing fault diagnosis under different working conditions is always a primary challenge. In this paper, a novel bearing fault diagnosis under different working conditions method is proposed based on domain adaptation using transferable features(DATF). The dataset of normal bearing and faulty bearings are obtained through the fast Fourier transformation(FFT) of raw vibration signals under different motor speeds and load conditions. Then we reduce marginal and conditional distributions simultaneously across domains based on maximum mean discrepancy(MMD) in feature space by refining pseudo test labels, which can be obtained by the Nearest-Neighbor(NN) classifier built on training data, and then a robust transferable feature representation for training and test domains is achieved after several iterations. With the help of the NN classifier trained on transferable features, bearing fault categories are identified accurately in final. Extensive experiment results show that the proposed method under different working conditions can identify the bearing faults accurately and outperforms obviously competitive approaches.

eess.SP↗

arXiv2018

Adversarial adaptive 1-D convolutional neural networks for bearing fault diagnosis under varying working condition

Traditional intelligent fault diagnosis of rolling bearings work well only under a common assumption that the labeled training data (source domain) and unlabeled testing data (target domain) are drawn from the same distribution. However, in many real-world applications, this assumption does not hold, especially when the working condition varies. In this paper, a new adversarial adaptive 1-D CNN called A2CNN is proposed to address this problem. A2CNN consists of four parts, namely, a source feature extractor, a target feature extractor, a label classifier and a domain discriminator. The layers between the source and target feature extractor are partially untied during the training stage to take both training efficiency and domain adaptation into consideration. Experiments show that A2CNN has strong fault-discriminative and domain-invariant capacity, and therefore can achieve high accuracy under different working conditions. We also visualize the learned features and the networks to explore the reasons behind the high performance of our proposed model.

eess.SP↗