arXiv subjects

Lei Wang

Publications and source records attributed to Lei Wang.

At least 1,045 records · Page 58Linked to original sources

arXiv2019

Hallucinating IDT Descriptors and I3D Optical Flow Features for Action Recognition with CNNs

In this paper, we revive the use of old-fashioned handcrafted video representations for action recognition and put new life into these techniques via a CNN-based hallucination step. Despite of the use of RGB and optical flow frames, the I3D model (amongst others) thrives on combining its output with the Improved Dense Trajectory (IDT) and extracted with its low-level video descriptors encoded via Bag-of-Words (BoW) and Fisher Vectors (FV). Such a fusion of CNNs and handcrafted representations is time-consuming due to pre-processing, descriptor extraction, encoding and tuning parameters. Thus, we propose an end-to-end trainable network with streams which learn the IDT-based BoW/FV representations at the training stage and are simple to integrate with the I3D model. Specifically, each stream takes I3D feature maps ahead of the last 1D conv. layer and learns to `translate' these maps to BoW/FV representations. Thus, our model can hallucinate and use such synthesized BoW/FV representations at the testing stage. We show that even features of the entire I3D optical flow stream can be hallucinated thus simplifying the pipeline. Our model saves 20-55h of computations and yields state-of-the-art results on four publicly available datasets.

cs.CV↗

arXiv2019

Variational Quantum Eigensolver with Fewer Qubits

We propose a qubit efficient scheme to study ground state properties of quantum many-body systems on near-term noisy intermediate scale quantum computers. One can obtain a tensor network representation of the ground state using a number of qubits smaller than the physical degrees of freedom. By increasing the qubits number, one can exponentially increase the bond dimension of the tensor network variational ansatz on a quantum computer. Moreover, we construct circuits blocks which respect U(1) and SU(2) symmetries of the physical system and show that they can significantly speed up the training process and alleviate the gradient vanishing problem. To demonstrate the feasibility of the qubit efficient variational quantum eigensolver in a practical setting, we perform first principle classical simulation of differentiable programming of the circuits. Using only $6$ qubits one can obtain the ground state of a $4\times 4$ square lattice frustrated Heisenberg model with fidelity over 97%. Arbitrarily long ranged correlations can also be measured on the same circuit after variational optimization.

quant-ph↗

arXiv2019

A Novel Unsupervised Camera-aware Domain Adaptation Framework for Person Re-identification

Unsupervised cross-domain person re-identification (Re-ID) faces two key issues. One is the data distribution discrepancy between source and target domains, and the other is the lack of labelling information in target domain. They are addressed in this paper from the perspective of representation learning. For the first issue, we highlight the presence of camera-level sub-domains as a unique characteristic of person Re-ID, and develop camera-aware domain adaptation to reduce the discrepancy not only between source and target domains but also across these sub-domains. For the second issue, we exploit the temporal continuity in each camera of target domain to create discriminative information. This is implemented by dynamically generating online triplets within each batch, in order to maximally take advantage of the steadily improved feature representation in training process. Together, the above two methods give rise to a novel unsupervised deep domain adaptation framework for person Re-ID. Experiments and ablation studies on benchmark datasets demonstrate its superiority and interesting properties.

cs.CV↗

arXiv2019

Theoretical Analysis for Extended Target Recovery in Randomized Stepped Frequency Radars

Randomized Stepped Frequency Radar (RSFR) is very attractive for tasks under complex electromagnetic environment. Due to the synthetic high range resolution in RSRFs, a target usually occupies a series of range cells and is called an extended target. To reconstruct the range-Doppler information in a RSFR, previous studies based on sparse recovery mainly exploit the sparsity of the target scene but do not adequately address the extended-target characteristics, which exist in many practical applications. Block sparsity, which combines the sparsity and the target extension, better characterizes a priori knowledge of the target scene in a wideband RSFR. This paper studies the RSFR range-Doppler reconstruction problem using block sparse recovery. Particularly, we theoretically analyze the block coherence and spectral norm of the observation matrix in RSFR and build a bound on the parameters of the radar, under which the exact recovery of the range-Doppler information is guaranteed. Both simulation and field experiment results demonstrate the superiority of the block sparse recovery over conventional sparse recovery in RSFRs.

eess.SP↗

arXiv2019

Edge AIBench: Towards Comprehensive End-to-end Edge Computing Benchmarking

In edge computing scenarios, the distribution of data and collaboration of workloads on different layers are serious concerns for performance, privacy, and security issues. So for edge computing benchmarking, we must take an end-to-end view, considering all three layers: client-side devices, edge computing layer, and cloud servers. Unfortunately, the previous work ignores this most important point. This paper presents the BenchCouncil's coordinated e ort on edge AI benchmarks, named Edge AIBench. In total, Edge AIBench models four typical application scenarios: ICU Patient Monitor, Surveillance Camera, Smart Home, and Autonomous Vehicle with the focus on data distribution and workload collaboration on three layers. Edge AIBench is a part of the open-source AIBench project, publicly available from http://www.benchcouncil.org/AIBench/index.html. We also build an edge computing testbed with a federated learning framework to resolve performance, privacy, and security issues.

cs.PF↗

arXiv2019

NPSA: Nonorthogonal Principal Skewness Analysis

Principal skewness analysis (PSA) has been introduced for feature extraction in hyperspectral imagery. As a third-order generalization of principal component analysis (PCA), its solution of searching for the locally maximum skewness direction is transformed into the problem of calculating the eigenpairs (the eigenvalues and the corresponding eigenvectors) of a coskewness tensor. By combining a fixed-point method with an orthogonal constraint, it can prevent the new eigenpairs from converging to the same maxima that has been determined before. However, the eigenvectors of the supersymmetric tensor are not inherently orthogonal in general, which implies that the results obtained by the search strategy used in PSA may unavoidably deviate from the actual eigenpairs. In this paper, we propose a new nonorthogonal search strategy to solve this problem and the new algorithm is named nonorthogonal principal skewness analysis (NPSA). The contribution of NPSA lies in the finding that the search space of the eigenvector to be determined can be enlarged by using the orthogonal complement of the Kronecker product of the previous one, instead of its orthogonal complement space. We give a detailed theoretical proof to illustrate why the new strategy can result in the more accurate eigenpairs. In addition, after some algebraic derivations, the complexity of the presented algorithm is also greatly reduced. Experiments with both simulated data and real multi/hyperspectral imagery demonstrate its validity in feature extraction.

cs.CV↗

arXiv2019

Intrinsic Picosecond Magnetic Switching Mechanism Assisted by an Electric Field in a Synthetic Antiferromagnetic Structure

The processional switching mechanism governs magnetic switching in magnetic tunnel junctions (MTJs) in the sub-nanosecond range, which limits the application of spin transfer torque magnetic random access memory (STT-MRAM) in the ultrafast region. In this paper, we propose a new picosecond magnetic switching mechanism in a synthetic antiferromagnetic (SAF) structure using the adjustable Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction controlled by an external electric field (E-field). It is shown that along with the sign change of the RKKY interaction in the SAF structure with an external E-field, the critical switching current density can be significantly reduced by one order of magnitude compared to that of a normal MTJ design at 100 ps; thus, this novel STT-MRAM can be written with a very low switching current density to avoid the MTJ breakdown problem and reduce the writing energy. To understand the physical origin of this abnormal phenomenon, a toy model is proposed in which the external-E-field-controlled sign change of the RKKY interaction in the SAF structure provides an extra contribution to the total energy that helps thespins overcome the energy barrier and break the processional switching mechanism.

cond-mat.mes-hall↗

arXiv2019

Topological Weyl-like Phonons and Nodal Line Phonons in Graphene

By means of first-principles calculations and modeling analysis, we have predicted that the traditional 2D-graphene hosts the topological phononic Weyl-like points (PWs) and phononic nodal line (PNL) in its phonon spectrum. The phonon dispersion of graphene hosts three type-I PWs (both PW1 and PW2 at the BZ corners \emph{K} and \emph{K}', and PW3 locating along the $Γ$-\emph{K} line), one type-II PW4 locating along the $Γ$-\emph{M} line, and one PNL surrounding the centered $Γ$ point in the $q_{x,y}$ plane. The calculations further reveal that Berry curvatures are vanishingly zero throughout the whole BZ, except for the positions of these four pairs of Weyl-like phonons, at which the non-zero singular Berry curvatures appear with the Berry phase of $π$ or -$π$, confirming its topological non-trivial nature. The topologically protected non-trivial phononic edge states have been also evidenced along both the zigzag-edged and armchair-edged boundaries. These results would pave the ways for further studies of topological phononic properties of graphene, such as phononic destructive interference with a suppression of backscattering and intrinsic phononic quantum Hall-like effects.

cond-mat.mes-hall↗

arXiv2019

Underlying topological Dirac nodal line mechanism of anomalously large electron-phonon coupling strength on Be (0001) surface

Beryllium was recently discovered to harbor a Dirac nodal line (DNL) in its bulk phase and the DNL-induced non-trivial drumhead-like surface states (DNSSs) on its (0001) surface, rationalizing several already-existing historic puzzles [Phys. Rev. Lett., \textbf{117}, 096401 (2016)]. However, to date the underlying mechanism, as to why its (0001) surface exhibits an anomalously large electron-phonon coupling effect ($λ_{e-ph}^s$ $\approx$ 1.0), remains unresolved. Here, by means of first-principles calculations we have evidenced that the coupling of the DNSSs with the phononic states mainly contributes to its novel surface \emph{e-ph} enhancement. Besides that the experimentally observed $λ_{e-ph}^s$ and the main Eliashberg coupling function (ECF) peaks have been reproduced well, we have decomposed the ECF, $α^{2}$$F$(\emph{k},\textbf{\emph{q}};\emph{v}), and the \emph{e-ph} coupling strength $λ(\emph{k},\textbf{\emph{q}};\emph{v})$ as a function of each electron momentum (\emph{k}), each phonon momentum (\textbf{\emph{q}}) and each phonon mode ($v$), evidencing the robust connection between the DNSSs and both $α^{2}$$F$(\emph{k},\textbf{\emph{q}};\emph{v}) and $λ(\emph{k},\textbf{\emph{q}};\emph{v})$. The results reveal the strong \emph{e-ph} coupling between the DNSSs and the phonon modes, which contributes over 80$\%$ of the $λ_{e-ph}^s$ coefficient on the Be (0001) surface. It highlights that the anomalously large \emph{e-ph} coefficient on the Be (0001) surface can be attributed to the presence of its DNL-induced DNSSs, clarifying the long-term debated mechanism.

cond-mat.mtrl-sci↗

arXiv2019

Energy-Efficient Power Control of Train-ground mmWave Communication for High Speed Trains

High speed train system has proven to be a very flexible and attractive system that can be developed under various circumstances and in different contexts and cultures. As a result, high speed trains are widely deployed around the world. Providing more reliable and higher data rate communication services for high speed trains has become one of the most urgent challenges. With vast amounts of spectrum available, the millimeter wave (mmWave) system is able to provide transmission rates of several gigabits per second for high speed trains. At the same time, mmWave communication also suffers from high attenuation, thus higher energy efficiency should be considered. This paper proposes an energy efficient power control scheme of train-ground mmWave communication for high speed trains. Considering a beam switching method for efficient beam alignment, we first establish position prediction model, the realistic direction antenna model and receive power model. And then we allocate the transmission power rationally through the power minimization algorithm while ensuring the total amount of transmission data. Based on this, this paper also develops a hybrid optimization scheme and finds the limit of total energy consumption when the number of segments goes to infinity. Through simulation with various system parameters and taking velocity estimation error into account, we demonstrate the superior performance of our schemes.

cs.NI↗

arXiv2019

Differentiable Programming Tensor Networks

Differentiable programming is a fresh programming paradigm which composes parameterized algorithmic components and trains them using automatic differentiation (AD). The concept emerges from deep learning but is not only limited to training neural networks. We present theory and practice of programming tensor network algorithms in a fully differentiable way. By formulating the tensor network algorithm as a computation graph, one can compute higher order derivatives of the program accurately and efficiently using AD. We present essential techniques to differentiate through the tensor networks contractions, including stable AD for tensor decomposition and efficient backpropagation through fixed point iterations. As a demonstration, we compute the specific heat of the Ising model directly by taking the second order derivative of the free energy obtained in the tensor renormalization group calculation. Next, we perform gradient based variational optimization of infinite projected entangled pair states for quantum antiferromagnetic Heisenberg model and obtain start-of-the-art variational energy and magnetization with moderate efforts. Differentiable programming removes laborious human efforts in deriving and implementing analytical gradients for tensor network programs, which opens the door to more innovations in tensor network algorithms and applications.

cond-mat.str-el↗

arXiv2019

Slater and Mott insulating states in the SU(6) Hubbard model

We perform large scale projector determinant quantum Monte-Carlo simulations to study the insulating states of the half-filled SU(6) Hubbard model on the square lattice. The transition from the antiferromagnetic state to the valence bond solid state occurs as increasing the Hubbard $U$. In contrast, in the SU(2) and SU(4) cases antiferromagnetism persists throughout the entire interaction range. In the SU(6) case, antiferromagnetism starts to develop in the weak interacting regime based on the Slater mechanism of Fermi surface nesting. As $U$ passes a crossover value $U^*/t\approx 9$, the single-particle gap scales linearly with $U$, marking the onset of Mott physics. In the Mott regime, antiferromagnetism becomes to be suppressed as $U$ increases, and vanishes after $U$ passes the critical value $U_{\rm AF,c}/t=13.3\pm 0.05$. The critical exponents are obtained via critical scalings as $ν_{\rm AF}=0.60\pm 0.02$ and $η_{\rm AF}=0.44\pm 0.03$. As $U$ further increases, the valence bond solid ordering appears exhibiting the anomalous dimension $η_{\rm VBS}=0.98\pm 0.01$.

cond-mat.str-el↗

arXiv2019

Evidence against a general positive eddy feedback in atmospheric blocking

The eddy straining mechanism of Shutts (1983; S83) has long been considered a main process for explaining the maintenance of atmospheric blocking. As hypothesized in S83, incoming synoptic eddies experience a meridional straining effect when approaching a split jetstream, and as a result, enhanced PV fluxes reinforce the block. A two-layer QG model is adopted here as a minimal model to conduct mechanism-denial experiments. While transient eddies' forcing is clearly critical to the formation and maintenance of a block, using a large ensemble, the authors demonstrate that the straining of generic eddies does not maintain blocks, thus challenge the idea of eddy straining serving as a positive feedback for the blocks. These results indicate that specific configurations of the eddy field are required for the maintenance stage. The authors also remark on the main supporting evidence in S83: the second-order induced flow is sensitive to the location of the wavemaker.

physics.ao-ph↗

arXiv2019

Loss Switching Fusion with Similarity Search for Video Classification

From video streaming to security and surveillance applications, video data play an important role in our daily living today. However, managing a large amount of video data and retrieving the most useful information for the user remain a challenging task. In this paper, we propose a novel video classification system that would benefit the scene understanding task. We define our classification problem as classifying background and foreground motions using the same feature representation for outdoor scenes. This means that the feature representation needs to be robust enough and adaptable to different classification tasks. We propose a lightweight Loss Switching Fusion Network (LSFNet) for the fusion of spatiotemporal descriptors and a similarity search scheme with soft voting to boost the classification performance. The proposed system has a variety of potential applications such as content-based video clustering, video filtering, etc. Evaluation results on two private industry datasets show that our system is robust in both classifying different background motions and detecting human motions from these background motions.

cs.CV↗

arXiv2019

A Comparative Review of Recent Kinect-based Action Recognition Algorithms

Video-based human action recognition is currently one of the most active research areas in computer vision. Various research studies indicate that the performance of action recognition is highly dependent on the type of features being extracted and how the actions are represented. Since the release of the Kinect camera, a large number of Kinect-based human action recognition techniques have been proposed in the literature. However, there still does not exist a thorough comparison of these Kinect-based techniques under the grouping of feature types, such as handcrafted versus deep learning features and depth-based versus skeleton-based features. In this paper, we analyze and compare ten recent Kinect-based algorithms for both cross-subject action recognition and cross-view action recognition using six benchmark datasets. In addition, we have implemented and improved some of these techniques and included their variants in the comparison. Our experiments show that the majority of methods perform better on cross-subject action recognition than cross-view action recognition, that skeleton-based features are more robust for cross-view recognition than depth-based features, and that deep learning features are suitable for large datasets.

cs.CV↗

arXiv2019

Ensemble Consider Kalman Filtering

In this paper, the ensemble consider Kalman filter is proposed to mitigate the negative effects of uncertain parameters in nonlinear dynamic and measurement models. The ensemble Kalman filter can avoid using the Jacobian matrices and reduce the computational complexity, the unknown parameters of the models still are not considered. By incorporating the statistics of the uncertain parameters into the state estimation formulations and using an augmented-state approach, the ensemble integration is reset by resampling the ensemble members in the new step, and the EnCKF algorithm is derived. Two numerical simulations show that the presented EnCKF can mitigate the negative effects of the uncertain parameters.

eess.SY↗

arXiv2019

Nonlinear elliptic equations on the upper half space

In this paper we shall classify all positive solutions of $ Δu =a u^p$ on the upper half space $ H =\Bbb{R}_+^n$ with nonlinear boundary condition $ {\partial u}/{\partial t}= - b u^q $ on $\partial H$ for both positive parameters $a, \ b>0$. We will prove that for $p \ge {(n+2)}/{(n-2)}, 1\leq q<{n}/{(n-2)}$ (and $n \ge 3$) all positive solutions are functions of last variable; for $p= {(n+2)}/{(n-2)}, q= {n}/{(n-2)}$ (and $n \ge 3$) positive solutions must be either some functions depending only on last variable, or radially symmetric functions.

math.AP↗

arXiv2019

A Novel Bi-hemispheric Discrepancy Model for EEG Emotion Recognition

The neuroscience study has revealed the discrepancy of emotion expression between left and right hemispheres of human brain. Inspired by this study, in this paper, we propose a novel bi-hemispheric discrepancy model (BiHDM) to learn the asymmetric differences between two hemispheres for electroencephalograph (EEG) emotion recognition. Concretely, we first employ four directed recurrent neural networks (RNNs) based on two spatial orientations to traverse electrode signals on two separate brain regions, which enables the model to obtain the deep representations of all the EEG electrodes' signals while keeping the intrinsic spatial dependence. Then we design a pairwise subnetwork to capture the discrepancy information between two hemispheres and extract higher-level features for final classification. Besides, in order to reduce the domain shift between training and testing data, we use a domain discriminator that adversarially induces the overall feature learning module to generate emotion-related but domain-invariant feature, which can further promote EEG emotion recognition. We conduct experiments on three public EEG emotional datasets, and the experiments show that the new state-of-the-art results can be achieved.

q-bio.NC↗