arXiv subjects

Zhe Wang

Publications and source records attributed to Zhe Wang.

At least 559 records · Page 31Linked to original sources

arXiv2019

Learning Depth-Guided Convolutions for Monocular 3D Object Detection

3D object detection from a single image without LiDAR is a challenging task due to the lack of accurate depth information. Conventional 2D convolutions are unsuitable for this task because they fail to capture local object and its scale information, which are vital for 3D object detection. To better represent 3D structure, prior arts typically transform depth maps estimated from 2D images into a pseudo-LiDAR representation, and then apply existing 3D point-cloud based object detectors. However, their results depend heavily on the accuracy of the estimated depth maps, resulting in suboptimal performance. In this work, instead of using pseudo-LiDAR representation, we improve the fundamental 2D fully convolutions by proposing a new local convolutional network (LCN), termed Depth-guided Dynamic-Depthwise-Dilated LCN (D$^4$LCN), where the filters and their receptive fields can be automatically learned from image-based depth maps, making different pixels of different images have different filters. D$^4$LCN overcomes the limitation of conventional 2D convolutions and narrows the gap between image representation and 3D point cloud representation. Extensive experiments show that D$^4$LCN outperforms existing works by large margins. For example, the relative improvement of D$^4$LCN against the state-of-the-art on KITTI is 9.1\% in the moderate setting. The code is available at https://github.com/dingmyu/D4LCN.

cs.CV↗

arXiv2019

Every Frame Counts: Joint Learning of Video Segmentation and Optical Flow

A major challenge for video semantic segmentation is the lack of labeled data. In most benchmark datasets, only one frame of a video clip is annotated, which makes most supervised methods fail to utilize information from the rest of the frames. To exploit the spatio-temporal information in videos, many previous works use pre-computed optical flows, which encode the temporal consistency to improve the video segmentation. However, the video segmentation and optical flow estimation are still considered as two separate tasks. In this paper, we propose a novel framework for joint video semantic segmentation and optical flow estimation. Semantic segmentation brings semantic information to handle occlusion for more robust optical flow estimation, while the non-occluded optical flow provides accurate pixel-level temporal correspondences to guarantee the temporal consistency of the segmentation. Moreover, our framework is able to utilize both labeled and unlabeled frames in the video through joint training, while no additional calculation is required in inference. Extensive experiments show that the proposed model makes the video semantic segmentation and optical flow estimation benefit from each other and outperforms existing methods under the same settings in both tasks.

cs.CV↗

arXiv2019

Determining the phase diagram of atomically thin layered antiferromagnet CrCl$_3$

Changes in the spin configuration of atomically-thin, magnetic van-der-Waals multilayers can cause drastic modifications in their opto-electronic properties. Conversely, the opto-electronic response of these systems provides information about the magnetic state, very difficult to obtain otherwise. Here we show that in CrCl$_3$ multilayers, the dependence of the tunnelling conductance on applied magnetic field ($H$), temperature ($T$), and number of layers ($N$) tracks the evolution of the magnetic state, enabling the magnetic phase diagram of these systems to be determined experimentally. Besides a high-field spin-flip transition occurring for all thicknesses, the in-plane magnetoconductance exhibits an even-odd effect due to a low-field spin-flop transition. If the layer number $N$ is even, the transition occurs at $μ_0 H \sim 0$ T due to the very small in-plane magnetic anisotropy, whereas for odd $N$ the net magnetization of the uncompensated layer causes the transition to occur at finite $H$. Through a quantitative analysis of the phenomena, we determine the interlayer exchange coupling as well as the staggered magnetization, and show that in CrCl$_3$ shape anisotropy dominates. Our results reveal the rich behaviour of atomically-thin layered antiferromagnets with weak magnetic anisotropy.

cond-mat.mes-hall↗

arXiv2019

A Driven Tagged Particle in Asymmetric Exclusion Processes

We consider the asymmetric exclusion process with a driven tagged particle on Z which has different jump rates from other particles and show that the tagged particle can have a ballistic behavior when the non-tagged particles have non-nearest-neighbor jump rates. We show the existence of some non-trivial invariant measures for the environment process viewed from the tagged particle. Our arguments are based on coupling, the martingale approach, and analyzing currents through fixed bonds.

math.PR↗

arXiv2019

A Driven Tagged Particle in Symmetric Exclusion Processes with Removals

We consider a driven tagged particle in a symmetric exclusion process on Z with a removal rule. In this process, untagged particles are removed once they jump to the left of the tagged particle. We investigate the behavior of the displacement of the tagged particle and prove limit theorems of it: an (annealed) law of large numbers, a central limit theorem, and a large deviation principle. We also characterize a class of ergodic measures for the environment process. Our approach is based on analyzing two auxiliary processes with associated martingales and a regenerative structure.

math.PR↗

arXiv2019

Non-perturbative high-harmonic generation in the three-dimensional Dirac semimetal Cd$_3$As$_2$

Harmonic generation is a general characteristic of driven nonlinear systems, and serves as an efficient tool for investigating the fundamental principles that govern the ultrafast nonlinear dynamics. In atomic gases, high-harmonic radiation is produced via a three-step process of ionization, acceleration, and recollision by strong-field infrared laser. This mechanism has been intensively investigated in the extreme ultraviolet and soft X-ray regions, forming the basis of attosecond research. In solid-state materials, which are characterized by crystalline symmetry and strong interactions, yielding of harmonics has just recently been reported. The observed high-harmonic generation was interpreted with fundamentally different mechanisms, such as interband tunneling combined with dynamical Bloch oscillations, intraband thermodynamics and nonlinear dynamics, and many-body electronic interactions. Here, in a distinctly different context of three-dimensional Dirac semimetal, we report on experimental observation of high-harmonic generation up to the seventh order driven by strong-field terahertz pulses. The observed non-perturbative high-harmonic generation is interpreted as a generic feature of terahertz-field driven nonlinear intraband kinetics of Dirac fermions. We anticipate that our results will trigger great interest in detection, manipulation, and coherent control of the nonlinear response in the vast family of three-dimensional Dirac and Weyl materials.

cond-mat.mtrl-sci↗

arXiv2019

Spin-flop transition in atomically thin MnPS$_3$ crystals

The magnetic state of atomically thin semiconducting layered antiferromagnets such as CrI$_3$ and CrCl$_3$ can be probed by forming tunnel barriers and measuring their resistance as a function of magnetic field ($H$) and temperature ($T$). This is possible because the tunneling magnetoresistance originates from a spin-filtering effect sensitive to the relative orientation of the magnetization in different layers, i.e., to the magnetic state of the multilayers. For systems in which antiferromagnetism occurs within an individual layer, however, no spin-filtering occurs: it is unclear whether this strategy can work. To address this issue, we investigate tunnel transport through atomically thin crystals of MnPS$_3$, a van der Waals semiconductor that in the bulk exhibits easy-axis antiferromagnetic order within the layers. For thick multilayers below $T\simeq 78$ K, a $T$-dependent magnetoresistance sets-in at $\sim 5$ T, and is found to track the boundary between the antiferromagnetic and the spin-flop phases known from bulk magnetization measurements. The magnetoresistance persists down to individual MnPS$_3$ monolayers with nearly unchanged characteristic temperature and magnetic field scales, albeit with a different dependence on $H$. We discuss the implications of these finding for the magnetic state of atomically thin MnPS$_3$ crystals, conclude that antiferromagnetic correlations persist down to the level of individual monolayers, and that tunneling magnetoresistance does allow magnetism in 2D insulating materials to be detected even in the absence of spin-filtering.

cond-mat.mes-hall↗

arXiv2019

Improved Zeroth-Order Variance Reduced Algorithms and Analysis for Nonconvex Optimization

Two types of zeroth-order stochastic algorithms have recently been designed for nonconvex optimization respectively based on the first-order techniques SVRG and SARAH/SPIDER. This paper addresses several important issues that are still open in these methods. First, all existing SVRG-type zeroth-order algorithms suffer from worse function query complexities than either zeroth-order gradient descent (ZO-GD) or stochastic gradient descent (ZO-SGD). In this paper, we propose a new algorithm ZO-SVRG-Coord-Rand and develop a new analysis for an existing ZO-SVRG-Coord algorithm proposed in Liu et al. 2018b, and show that both ZO-SVRG-Coord-Rand and ZO-SVRG-Coord (under our new analysis) outperform other exiting SVRG-type zeroth-order methods as well as ZO-GD and ZO-SGD. Second, the existing SPIDER-type algorithm SPIDER-SZO (Fang et al. 2018) has superior theoretical performance, but suffers from the generation of a large number of Gaussian random variables as well as a $\sqrtε$-level stepsize in practice. In this paper, we develop a new algorithm ZO-SPIDER-Coord, which is free from Gaussian variable generation and allows a large constant stepsize while maintaining the same convergence rate and query complexity, and we further show that ZO-SPIDER-Coord automatically achieves a linear convergence rate as the iterate enters into a local PL region without restart and algorithmic modification.

cs.LG↗

arXiv2019

RiWalk: Fast Structural Node Embedding via Role Identification

Nodes performing different functions in a network have different roles, and these roles can be gleaned from the structure of the network. Learning latent representations for the roles of nodes helps to understand the network and to transfer knowledge across networks. However, most existing structural embedding approaches suffer from high computation and space cost or rely on heuristic feature engineering. Here we propose RiWalk, a flexible paradigm for learning structural node representations. It decouples the structural embedding problem into a role identification procedure and a network embedding procedure. Through role identification, rooted kernels with structural dependencies kept are built to better integrate network embedding methods. To demonstrate the effectiveness of RiWalk, we develop two different role identification methods named RiWalk-SP and RiWalk-WL respectively and employ random walk based network embedding methods. Experiments on within-network classification tasks show that our proposed algorithms achieve comparable performance with other baselines while being an order of magnitude more efficient. Besides, we also conduct across-network role classification tasks. The results show potential of structural embeddings in transfer learning. RiWalk is also scalable, making it capable of capturing structural roles in massive networks.

cs.SI↗

arXiv2019

A compact and efficient three-dimensional microfluidic mixer

Microfluidic mixing is a fundamental functionality in most lab on a chip (LOC) systems,whereas realization of efficient mixing is challenging in microfluidic channels due to the small Reynolds numbers. Here, we design and fabricate a compact three-dimensional (3D) micromixer to enable efficient mixing at various flow rates. The performance of the fabricated micromixer was examined using blue and red inks. The extreme flexibility in fabricating microfluidic structures of arbitrary 3D geometries using femtosecond laser micromachining allows us to tackle the major disadvantageous effects for optimizing the mixing efficiency.

physics.app-ph↗

arXiv2019

One-dimensional Quantum Spin Dynamics of Bethe String States

Quantum dynamics of strongly correlated systems is a challenging problem. Although the low energy fractional excitations of one dimensional integrable models are often well-understood, exploring quantum dynamics in these systems remains challenging in the gapless regime, especially at intermediate and high energies. Based on the algebraic Bethe ansatz formalism, we study spin dynamics in a representative one dimensional strongly correlated model, {\it i.e. }, the antiferromagnetic spin-$\frac{1}{2}$ XXZ chain with the Ising anisotropy, via the form-factor formulae. Various excitations at different energy scales are identified crucial to the dynamic spin structure factors under the guidance of sum rules. At small magnetic polarizations, gapless excitations dominate the low energy spin dynamics arising from the magnetic-field-induced incommensurability. In contrast, spin dynamics at intermediate and high energies is characterized by the two- and three-string states, which are multi-particle excitations based on the commensurate Néel ordered background. Our work is helpful for experimental studies on spin dynamics in both condensed matter and cold atom systems beyond the low energy effective Luttinger liquid theory. Based on an intuitive physical picture, we speculate that the dynamic feature at high energies due to the multi-particle anti-bound state excitations can be generalized to non-integrable spin systems.

cond-mat.str-el↗

arXiv2019

Variable Population Memetic Search: A Case Study on the Critical Node Problem

Population-based memetic algorithms have been successfully applied to solve many difficult combinatorial problems. Often, a population of fixed size was used in such algorithms to record some best solutions sampled during the search. However, given the particular features of the problem instance under consideration, a population of variable size would be more suitable to ensure the best search performance possible. In this work, we propose variable population memetic search (VPMS), where a strategic population sizing mechanism is used to dynamically adjust the population size during the memetic search process. Our VPMS approach starts its search from a small population of only two solutions to focus on exploitation, and then adapts the population size according to the search status to continuously influence the balancing between exploitation and exploration. We illustrate an application of the VPMS approach to solve the challenging critical node problem (CNP). We show that the VPMS algorithm integrating a variable population, an effective local optimization procedure (called diversified late acceptance search) and a backbone-based crossover operator performs very well compared to state-of-the-art CNP algorithms. The algorithm is able to discover new upper bounds for 13 instances out of the 42 popular benchmark instances, while matching 23 previous best-known upper bounds.

cs.NE↗

arXiv2019

Dynamic Equivalence between Soft Star Polymers and Hard Spheres

Understanding the dynamics of soft colloids, such as star polymers, dendrimers, and microgels, is of scientific and practical importance. It is known that the excluded volume effect plays a key role in colloidal dynamics. Here, we propose a condition of compressibility equivalence that provides a simple method to experimentally evaluate the excluded volume of soft colloids from a thermodynamic view. We apply this condition to survey the dynamics of a series of star polymer dispersions. It is found that as the concentration increases, the slowing of the long-time self-diffusivity of the star polymer, normalized by the short-time self-diffusivity, can be mapped onto the hard-sphere behavior. This phenomenon reveals the dynamic equivalence between soft colloids and hard spheres, despite the apparent complexity of the interparticle interaction of the soft colloids. The methods for measuring the osmotic compressibility and the self-diffusivities of soft colloidal dispersions are also presented.

cond-mat.soft↗

arXiv2019

Robust Multi-Modality Multi-Object Tracking

Multi-sensor perception is crucial to ensure the reliability and accuracy in autonomous driving system, while multi-object tracking (MOT) improves that by tracing sequential movement of dynamic objects. Most current approaches for multi-sensor multi-object tracking are either lack of reliability by tightly relying on a single input source (e.g., center camera), or not accurate enough by fusing the results from multiple sensors in post processing without fully exploiting the inherent information. In this study, we design a generic sensor-agnostic multi-modality MOT framework (mmMOT), where each modality (i.e., sensors) is capable of performing its role independently to preserve reliability, and further improving its accuracy through a novel multi-modality fusion module. Our mmMOT can be trained in an end-to-end manner, enables joint optimization for the base feature extractor of each modality and an adjacency estimator for cross modality. Our mmMOT also makes the first attempt to encode deep representation of point cloud in data association process in MOT. We conduct extensive experiments to evaluate the effectiveness of the proposed framework on the challenging KITTI benchmark and report state-of-the-art performance. Code and models are available at https://github.com/ZwwWayne/mmMOT.

cs.CV↗

arXiv2019

Two-dimensional Ferromagnetic van der Waals CrX3 (X=Cl, Br, I) Monolayers with Enhanced Anisotropy and Curie Temperature

Among the recently widely studied van der Waals layered magnets CrX3 (X=Cl, Br, I), CrCl3 monolayer (ML) is particularly puzzling as it is solely shown by experiments to have an in-plane magnetic easy axis and, furthermore, all of previous first-principles calculation results contradict this. Through systematical first-principles calculations,we unveil that its in-plane shape anisotropy that dominates over its weak perpendicular magnetocrystalline anisotropy is responsible for the in-plane magnetic easy axis of CrCl3 ML. To tune the in-plane ferromagnetism of CrCl3 ML into the desirable perpendicular one, we propose substituting Cr with isovalent tungsten (W). We find that CrWCl6 has a strong perpendicular magnetic anisotropy and a high Curie temperature up to 76 K. Our work not only gives insight into understanding the two-dimensional ferromagnetism of van der Waals MLs but also sheds new light on engineering their performances for nanodevices.

cond-mat.mtrl-sci↗

arXiv2019

Dispersions of Many-Body Bethe strings

Complex bound states of magnetic excitations, known as Bethe string, were predicted almost a century ago to exist in one-dimensional quantum magnets 1. The dispersions of the string states have so far remained the subject of intensive theoretical studies 2-7. By performing neutron scattering experiments on the one-dimensional Heisenberg-Ising antiferromagnet SrCo2V2O8 in high longitudinal magnetic fields, we reveal in detail the dispersion relations of the string states over the full Brillouin zone, as well as their magnetic field dependences. Furthermore the characteristic energy, the scattering intensity and linewidth of the observed string states exhibit excellent agreement with our precise Bethe Ansatz calculations. Our results establish the important role of string states in the quantum spin dynamics of one-dimensional systems, and will invoke studies of their dynamical properties in more general many-body systems.

cond-mat.str-el↗

arXiv2019

High-Field Quantum Disordered State in $α$-RuCl3: Spin Flips, Bound States, and a Multi-Particle Continuum

Layered $α$-RuCl3 has been discussed as a proximate Kitaev spin liquid compound. Raman and THz spectroscopy of magnetic excitations confirm that the low-temperature antiferromagnetic ordered phase features a broad Raman continuum, together with two magnon-like excitations at 2.7 and 3.6 meV, respectively. The continuum strength is maximized as long-range order is suppressed by an external magnetic field. The state above the field-induced quantum phase transition around 7.5 T is characterized by a gapped multi-particle continuum out of which a two-particle bound state emerges, together with a well-defined single-particle excitation at lower energy. Exact diagonalization calculations demonstrate that Kitaev and off-diagonal exchange terms in the Fleury-Loudon operator are crucial for the occurrence of these features in the Raman spectra. Our study firmly establishes the partially-polarized quantum disordered character of the high-field phase.

cond-mat.str-el↗

arXiv2019

Low-temperature monoclinic layer stacking in atomically thin CrI$_3$ crystals

Chromium triiodide, CrI$_3$, is emerging as a promising magnetic two-dimensional semiconductor where spins are ferromagnetically aligned within a single layer. Potential applications in spintronics arise from an antiferromagnetic ordering between adjacent layers that gives rise to spin filtering and a large magnetoresistance in tunnelling devices. This key feature appears only in thin multilayers and it is not inherited from bulk crystals, where instead neighbouring layers share the same ferromagnetic spin orientation. This discrepancy between bulk and thin samples is unexpected, as magnetic ordering between layers arises from exchange interactions that are local in nature and should not depend strongly on thickness. Here we solve this controversy and show through polarization resolved Raman spectroscopy that thin multilayers do not undergo a structural phase transition typical of bulk crystals. As a consequence, a different stacking pattern is present in thin and bulk samples at the temperatures at which magnetism sets in and, according to previous first-principles simulations, this results in a different interlayer magnetic ordering. Our experimental findings provide evidence for the strong interplay between stacking order and magnetism in CrI$_3$, opening interesting perspectives to design the magnetic state of van der Waals multilayers.

cond-mat.mes-hall↗