arXiv subjects

Wei Hu

Publications and source records attributed to Wei Hu.

At least 307 records · Page 17Linked to original sources

arXiv2019

Learning to Exploit Long-term Relational Dependencies in Knowledge Graphs

We study the problem of knowledge graph (KG) embedding. A widely-established assumption to this problem is that similar entities are likely to have similar relational roles. However, existing related methods derive KG embeddings mainly based on triple-level learning, which lack the capability of capturing long-term relational dependencies of entities. Moreover, triple-level learning is insufficient for the propagation of semantic information among entities, especially for the case of cross-KG embedding. In this paper, we propose recurrent skipping networks (RSNs), which employ a skipping mechanism to bridge the gaps between entities. RSNs integrate recurrent neural networks (RNNs) with residual learning to efficiently capture the long-term relational dependencies within and between KGs. We design an end-to-end framework to support RSNs on different tasks. Our experimental results showed that RSNs outperformed state-of-the-art embedding-based methods for entity alignment and achieved competitive performance for KG completion.

cs.AI↗

arXiv2019

Matrix Linear Discriminant Analysis

We propose a novel linear discriminant analysis approach for the classification of high-dimensional matrix-valued data that commonly arises from imaging studies. Motivated by the equivalence of the conventional linear discriminant analysis and the ordinary least squares, we consider an efficient nuclear norm penalized regression that encourages a low-rank structure. Theoretical properties including a non-asymptotic risk bound and a rank consistency result are established. Simulation studies and an application to electroencephalography data show the superior performance of the proposed method over the existing approaches.

stat.ME↗

arXiv2019

Optimized Skeleton-based Action Recognition via Sparsified Graph Regression

With the prevalence of accessible depth sensors, dynamic human body skeletons have attracted much attention as a robust modality for action recognition. Previous methods model skeletons based on RNN or CNN, which has limited expressive power for irregular skeleton joints. While graph convolutional networks (GCN) have been proposed to address irregular graph-structured data, the fundamental graph construction remains challenging. In this paper, we represent skeletons naturally on graphs, and propose a graph regression based GCN (GR-GCN) for skeleton-based action recognition, aiming to capture the spatio-temporal variation in the data. As the graph representation is crucial to graph convolution, we first propose graph regression to statistically learn the underlying graph from multiple observations. In particular, we provide spatio-temporal modeling of skeletons and pose an optimization problem on the graph structure over consecutive frames, which enforces the sparsity of the underlying graph for efficient representation. The optimized graph not only connects each joint to its neighboring joints in the same frame strongly or weakly, but also links with relevant joints in the previous and subsequent frames. We then feed the optimized graph into the GCN along with the coordinates of the skeleton sequence for feature learning, where we deploy high-order and fast Chebyshev approximation of spectral graph convolution. Further, we provide analysis of the variation characterization by the Chebyshev approximation. Experimental results validate the effectiveness of the proposed graph regression and show that the proposed GR-GCN achieves the state-of-the-art performance on the widely used NTU RGB+D, UT-Kinect and SYSU 3D datasets.

cs.CV↗

arXiv2019

Noise-Tolerant Paradigm for Training Face Recognition CNNs

Benefit from large-scale training datasets, deep Convolutional Neural Networks(CNNs) have achieved impressive results in face recognition(FR). However, tremendous scale of datasets inevitably lead to noisy data, which obviously reduce the performance of the trained CNN models. Kicking out wrong labels from large-scale FR datasets is still very expensive, although some cleaning approaches are proposed. According to the analysis of the whole process of training CNN models supervised by angular margin based loss(AM-Loss) functions, we find that the $θ$ distribution of training samples implicitly reflects their probability of being clean. Thus, we propose a novel training paradigm that employs the idea of weighting samples based on the above probability. Without any prior knowledge of noise, we can train high performance CNN models with large-scale FR datasets. Experiments demonstrate the effectiveness of our training paradigm. The codes are available at https://github.com/huangyangyu/NoiseFace.

cs.CV↗

arXiv2019

High-throughput investigation of tunable superconductivity in FeSe films

There is an ongoing debate about the relative importance of structural change versus doping charge carriers on the mechanism of superconductivity in Fe-based materials. Elucidating this issue is a major challenge since it would require a large number of samples where structure properties or the carrier density is systematically varied. FeSe, with its structural simplicity, is an ideal platform for addressing this question. It has been demonstrated that the superconductivity in this material can be controlled through crystal lattice tuning, as well as electronic structure manipulation. Here, we apply a high-throughput methodology to FeSe to systematically delineate the interdependence of its structural and electronic properties. Using a dual-beam pulsed laser deposition, we have generated FeSe films with a marked gradient in the superconducting transition temperature (below 2 K < Tc < 12 K) across 1 cm width of the films. The Tc gradient films display ~ 1% continuous stretch and compression in the out-of-plane and in-plane lattice constants respectively, triggering the continuous enhancement of superconductivity. Combining transport and angular-resolved photoemission measurements on uniform FeSe films with tunable Tc from 3 K to 14 K, we find that the electron carrier density is intimately correlated with Tc, i.e., it increases by a factor of 6 and ultimately surpasses the almost constant hole concentration. Our transmission electron microscope and band structure calculations reveal that rather than by shifting the chemical potential, the enhanced superconductivity is linked to the selective adjustment of the dxy band dispersion across the Fermi level by means of reduced local lattice distortions. Therefore, such novel mechanism provides a key to understand discrete superconducting phases in FeSe.

cond-mat.supr-con↗

arXiv2019

Nearly Optimal Dynamic $k$-Means Clustering for High-Dimensional Data

We consider the $k$-means clustering problem in the dynamic streaming setting, where points from a discrete Euclidean space $\{1, 2, \ldots, Δ\}^d$ can be dynamically inserted to or deleted from the dataset. For this problem, we provide a one-pass coreset construction algorithm using space $\tilde{O}(k\cdot \mathrm{poly}(d, \logΔ))$, where $k$ is the target number of centers. To our knowledge, this is the first dynamic geometric data stream algorithm for $k$-means using space polynomial in dimension and nearly optimal (linear) in $k$.

cs.DS↗

arXiv2019

Linear Convergence of the Primal-Dual Gradient Method for Convex-Concave Saddle Point Problems without Strong Convexity

We consider the convex-concave saddle point problem $\min_{x}\max_{y} f(x)+y^\top A x-g(y)$ where $f$ is smooth and convex and $g$ is smooth and strongly convex. We prove that if the coupling matrix $A$ has full column rank, the vanilla primal-dual gradient method can achieve linear convergence even if $f$ is not strongly convex. Our result generalizes previous work which either requires $f$ and $g$ to be quadratic functions or requires proximal mappings for both $f$ and $g$. We adopt a novel analysis technique that in each iteration uses a "ghost" update as a reference, and show that the iterates in the primal-dual gradient method converge to this "ghost" sequence. Using the same technique we further give an analysis for the primal-dual stochastic variance reduced gradient (SVRG) method for convex-concave saddle point problems with a finite-sum structure.

math.OC↗

arXiv2019

A spatially adaptive high-order meshless method for fluid-structure interactions

We present a scheme implementing an a posteriori refinement strategy in the context of a high-order meshless method for problems involving point singularities and fluid-solid interfaces. The generalized moving least squares (GMLS) discretization used in this work has been previously demonstrated to provide high-order compatible discretization of the Stokes and Darcy problems, offering a high-fidelity simulation tool for problems with moving boundaries. The meshless nature of the discretization is particularly attractive for adaptive h-refinement, especially when resolving the near-field aspects of variables and point singularities governing lubrication effects in fluid-structure interactions. We demonstrate that the resulting spatially adaptive GMLS method is able to achieve optimal convergence in the presence of singularities for both the div-grad and Stokes problems. Further, we present a series of simulations for flows of colloid suspensions, in which the refinement strategy efficiently achieved highly accurate solutions, particularly for colloids with complex geometries.

physics.comp-ph↗

arXiv2019

Anomalies of upper critical field in the spinel superconductor LiTi$_2$O$_{4-δ}$

High-field electrical transport and point-contact tunneling spectroscopy were used to investigate superconducting properties of the unique spinel oxide, LiTi$_2$O$_{4-δ}$ films with various oxygen content. We find that the upper critical field $B_\mathrm{c2}$ gradually increases as more oxygen impurities are brought into the samples by carefully tuning the deposition atmosphere. It is striking that although the superconducting transition temperature and energy gap are almost unchanged, an astonishing isotropic $B_\mathrm{c2}$ up to $\sim$ 26 Tesla is observed in oxygen-rich sample, which is doubled compared to the anoxic sample and breaks the Pauli limit. Such anomalies of $B_\mathrm{c2}$ were rarely reported in other three dimensional superconductors. Combined with all the anomalies, three dimensional spin-orbit interaction induced by tiny oxygen impurities is naturally proposed to account for the remarkable enhancement of $B_\mathrm{c2}$ in oxygen-rich LiTi$_2$O$_{4-δ}$ films. Such mechanism could be general and therefore provides ideas for optimizing practical superconductors with higher $B_\mathrm{c2}$.

cond-mat.supr-con↗

arXiv2018

DSKG: A Deep Sequential Model for Knowledge Graph Completion

Knowledge graph (KG) completion aims to fill the missing facts in a KG, where a fact is represented as a triple in the form of $(subject, relation, object)$. Current KG completion models compel two-thirds of a triple provided (e.g., $subject$ and $relation$) to predict the remaining one. In this paper, we propose a new model, which uses a KG-specific multi-layer recurrent neural network (RNN) to model triples in a KG as sequences. It outperformed several state-of-the-art KG completion models on the conventional entity prediction task for many evaluation metrics, based on two benchmark datasets and a more difficult dataset. Furthermore, our model is enabled by the sequential characteristic and thus capable of predicting the whole triples only given one entity. Our experiments demonstrated that our model achieved promising performance on this new triple prediction task.

cs.LG↗

arXiv2018

Feature Preserving and Uniformity-controllable Point Cloud Simplification on Graph

With the development of 3D sensing technologies, point clouds have attracted increasing attention in a variety of applications for 3D object representation, such as autonomous driving, 3D immersive tele-presence and heritage reconstruction. However, it is challenging to process large-scale point clouds in terms of both computation time and storage due to the tremendous amounts of data. Hence, we propose a point cloud simplification algorithm, aiming to strike a balance between preserving sharp features and keeping uniform density during resampling. In particular, leveraging on graph spectral processing, we represent irregular point clouds naturally on graphs, and propose concise formulations of feature preservation and density uniformity based on graph filters. The problem of point cloud simplification is finally formulated as a trade-off between the two factors and efficiently solved by our proposed algorithm. Experimental results demonstrate the superiority of our method, as well as its efficient application in point cloud registration.

cs.CV↗

arXiv2018

Exploring Hypergraph Representation on Face Anti-spoofing Beyond 2D Attacks

Face anti-spoofing plays a crucial role in protecting face recognition systems from various attacks. Previous model-based and deep learning approaches achieve satisfactory performance for 2D face spoofs, but remain limited for more advanced 3D attacks such as vivid masks. In this paper, we address 3D face anti-spoofing via the proposed Hypergraph Convolutional Neural Networks (HGCNN). Firstly, we construct a computation-efficient and posture-invariant face representation with only a few key points on hypergraphs. The hypergraph representation is then fed into the designed HGCNN with hypergraph convolution for feature extraction, while the depth auxiliary is also exploited for 3D mask anti-spoofing. Further, we build a 3D face attack database with color, depth and infrared light information to overcome the deficiency of 3D face anti-spoofing data. Experiments show that our method achieves the state-of-the-art performance over widely used 3D and 2D databases as well as the proposed one under various tests.

cs.CV↗

arXiv2018

The enhanced X-ray Timing and Polarimetry mission - eXTP

In this paper we present the enhanced X-ray Timing and Polarimetry mission - eXTP. eXTP is a space science mission designed to study fundamental physics under extreme conditions of density, gravity and magnetism. The mission aims at determining the equation of state of matter at supra-nuclear density, measuring effects of QED, and understanding the dynamics of matter in strong-field gravity. In addition to investigating fundamental physics, eXTP will be a very powerful observatory for astrophysics that will provide observations of unprecedented quality on a variety of galactic and extragalactic objects. In particular, its wide field monitoring capabilities will be highly instrumental to detect the electro-magnetic counterparts of gravitational wave sources. The paper provides a detailed description of: (1) the technological and technical aspects, and the expected performance of the instruments of the scientific payload; (2) the elements and functions of the mission, from the spacecraft to the ground segment.

astro-ph.IM↗

arXiv2018

An HAM-Based Analytic Modeling Methodology for Memristor Enabling Fast Convergence

Memristor has great application prospects in various high-performance electronic systems, such as memory, artificial intelligence, and neural networks, due to its fast speed, nano-scale dimensions, and low-power consumption. However, traditional nonanalytic models and lately reported analytic models for memristor have the problems of nonconvergence and slow convergence, respectively. These problems lay great obstacles in the analysis, simulation and design of memristor. To address these problems, a modeling methodology for analytic approximate solution of the state variable in memristor is proposed in this work. This methodology solves the governing equation of memristor by adopting the Homotopy Analysis Method (HAM) for the first time, and the convergence performance of the methodology is enhanced by an optimized convergence-control parameter in the HAM. The simulation results, compared with the reported analytic models, demonstrate that the HAM-based modeling methodology achieves faster convergence while guaranteeing sufficient accuracy. Based on the methodology, it can be simultaneously revealed that highly nonlinearity is the potential sources of slow convergence in analytic models, which is beneficial for the analysis and design guidance. In addition, a Spice subcircuit is constructed based on the obtained HAM model, and then it is integrated into an oscillator to verify its applicability. Due to the generality of HAM, this methodology may be easily extended to other memory devices.

physics.app-ph↗

arXiv2018

Evolution of orbital angular momentum in a soft quasi-periodic structure with topological defects

We propose a quasi-periodic structure (QPS) with topological defects. The analytical expression of the corresponding Fourier spectrum is derived, which reflects the combined effects of topological structure and quasitranslational symmetry. Light-matter interaction therein brings unusual diffraction characteristics with exotic evolution of orbital angular momentum (OAM). Long-range correlation of QPS resulted in multi-fractal and pairwise distribution of optical singularities. A general conversation law of OAM was revealed. A liquid crystal photopatterning QPS is fabricated to demonstrate the above characteristics. Dynamic reconfigurable manipulation of optical singularities was achieved. Our approach offers the opportunity to manipulate OAM with multiple degrees of freedom, which has promising applications in multi-channel quantum information processing and highdimensional quantum state generation.

physics.optics↗

arXiv2018

Recurrent Skipping Networks for Entity Alignment

We consider the problem of learning knowledge graph (KG) embeddings for entity alignment (EA). Current methods use the embedding models mainly focusing on triple-level learning, which lacks the ability of capturing long-term dependencies existing in KGs. Consequently, the embedding-based EA methods heavily rely on the amount of prior (known) alignment, due to the identity information in the prior alignment cannot be efficiently propagated from one KG to another. In this paper, we propose RSN4EA (recurrent skipping networks for EA), which leverages biased random walk sampling for generating long paths across KGs and models the paths with a novel recurrent skipping network (RSN). RSN integrates the conventional recurrent neural network (RNN) with residual learning and can largely improve the convergence speed and performance with only a few more parameters. We evaluated RSN4EA on a series of datasets constructed from real-world KGs. Our experimental results showed that it outperformed a number of state-of-the-art embedding-based EA methods and also achieved comparable performance for KG completion.

cs.CL↗

arXiv2018

Algorithmic Regularization in Learning Deep Homogeneous Models: Layers are Automatically Balanced

We study the implicit regularization imposed by gradient descent for learning multi-layer homogeneous functions including feed-forward fully connected and convolutional deep neural networks with linear, ReLU or Leaky ReLU activation. We rigorously prove that gradient flow (i.e. gradient descent with infinitesimal step size) effectively enforces the differences between squared norms across different layers to remain invariant without any explicit regularization. This result implies that if the weights are initially small, gradient flow automatically balances the magnitudes of all layers. Using a discretization argument, we analyze gradient descent with positive step size for the non-convex low-rank asymmetric matrix factorization problem without any regularization. Inspired by our findings for gradient flow, we prove that gradient descent with step sizes $η_t = O\left(t^{-\left( \frac12+δ\right)} \right)$ ($0<δ\le\frac12$) automatically balances two low-rank factors and converges to a bounded global optimum. Furthermore, for rank-$1$ asymmetric matrix factorization we give a finer analysis showing gradient descent with constant step size converges to the global minimum at a globally linear rate. We believe that the idea of examining the invariance imposed by first order algorithms in learning homogeneous models could serve as a fundamental building block for studying optimization for learning deep models.

cs.LG↗

arXiv2018

Light Reconfigurable Geometric Phase Optical Element with Multi-stable States

We present the design methodology of a light reconfigurable geometric phase optical element with multi-stable diffraction efficiency states, enabled by a photoresponsive self-organized chiral liquid crystal. Experimental demonstration shows the device exhibits a broad diffraction efficiency tunable range that can be smoothly modulated under alternate stimulation of ultraviolet and green lights. Distinctive to previous designs, the regulation of diffraction efficiency fundamentally stems from the modulation of geometric phase together with dynamical phase retardation, and any intermediate diffractive state is memorized. Such multi-stability facilitates applications including energy-saving all-optical signal processing in classical and quantum level, and phase hologram for anti-counterfeit.

physics.optics↗