arXiv subjects

Xiaotian Zhang

Publications and source records attributed to Xiaotian Zhang.

At least 55 records · Page 3Linked to original sources

arXiv2023

Investigating Glyph Phonetic Information for Chinese Spell Checking: What Works and What's Next

While pre-trained Chinese language models have demonstrated impressive performance on a wide range of NLP tasks, the Chinese Spell Checking (CSC) task remains a challenge. Previous research has explored using information such as glyphs and phonetics to improve the ability to distinguish misspelled characters, with good results. However, the generalization ability of these models is not well understood: it is unclear whether they incorporate glyph-phonetic information and, if so, whether this information is fully utilized. In this paper, we aim to better understand the role of glyph-phonetic information in the CSC task and suggest directions for improvement. Additionally, we propose a new, more challenging, and practical setting for testing the generalizability of CSC models. All code is made publicly available.

cs.CL↗

arXiv2023

Learning Global-Local Correspondence with Semantic Bottleneck for Logical Anomaly Detection

This paper presents a novel framework, named Global-Local Correspondence Framework (GLCF), for visual anomaly detection with logical constraints. Visual anomaly detection has become an active research area in various real-world applications, such as industrial anomaly detection and medical disease diagnosis. However, most existing methods focus on identifying local structural degeneration anomalies and often fail to detect high-level functional anomalies that involve logical constraints. To address this issue, we propose a two-branch approach that consists of a local branch for detecting structural anomalies and a global branch for detecting logical anomalies. To facilitate local-global feature correspondence, we introduce a novel semantic bottleneck enabled by the visual Transformer. Moreover, we develop feature estimation networks for each branch separately to detect anomalies. Our proposed framework is validated using various benchmarks, including industrial datasets, Mvtec AD, Mvtec Loco AD, and the Retinal-OCT medical dataset. Experimental results show that our method outperforms existing methods, particularly in detecting logical anomalies.

cs.CV↗

arXiv2022

SDCL: Self-Distillation Contrastive Learning for Chinese Spell Checking

Due to the ambiguity of homophones, Chinese Spell Checking (CSC) has widespread applications. Existing systems typically utilize BERT for text encoding. However, CSC requires the model to account for both phonetic and graphemic information. To adapt BERT to the CSC task, we propose a token-level self-distillation contrastive learning method. We employ BERT to encode both the corrupted and corresponding correct sentence. Then, we use contrastive learning loss to regularize corrupted tokens' hidden states to be closer to counterparts in the correct sentence. On three CSC datasets, we confirmed our method provides a significant improvement above baselines.

cs.CL↗

arXiv2021

Friction modulation in limbless, three-dimensional gaits and heterogeneous terrains

Motivated by a possible convergence of terrestrial limbless locomotion strategies ultimately determined by interfacial effects, we show how both 3D gait alterations and locomotory adaptations to heterogeneous terrains can be understood through the lens of local friction modulation. Via an `everything-is-friction' modeling approach, compounded by 3D simulations, the emergence and disappearance of a range of locomotory behaviors observed in nature is systematically explained in relation to inhabited environments. Our approach also simplifies the treatment of terrain heterogeneity, whereby even solid obstacles may be seen as high friction regions, which we confirm against experiments of snakes `diffracting' while traversing rows of posts [1], similar to optical waves. We further this optic analogy by illustrating snake refraction, reflection and lens focusing. We use these insights to engineer surface friction patterns and demonstrate passive snake navigation in complex topographies. Overall, our study outlines a unified view that connects active and passive 3D mechanics with heterogeneous interfacial effects to explain a broad set of biological observations, and potentially inspire engineering design.

physics.bio-ph↗

arXiv2021

Opportunities in Electrically Tunable 2D Materials Beyond Graphene: Recent Progress and Future Outlook

The interest in two-dimensional and layered materials continues to expand, driven by the compelling properties of individual atomic layers that can be stacked and/or twisted into synthetic heterostructures. The plethora of electronic properties as well as the emergence of many different quasiparticles, including plasmons, polaritons, trions and excitons with large, tunable binding energies that all can be controlled and modulated through electrical means has given rise to many device applications. In addition, these materials exhibit both room-temperature spin and valley polarization, magnetism, superconductivity, piezoelectricity that are intricately dependent on the composition, crystal structure, stacking, twist angle, layer number and phases of these materials. Initial results on graphene exfoliated from single bulk crystals motivated the development of wide-area, high purity synthesis and heterojunctions with atomically clean interfaces. Now by opening this design space to new synthetic two-dimensional materials "beyond graphene", it is possible to explore uncharted opportunities in designing novel heterostructures for electrical tunable devices. To fully reveal the emerging functionalities and opportunities of these atomically thin materials in practical applications, this review highlights several representative and noteworthy research directions in the use of electrical means to tune these aforementioned physical and structural properties, with an emphasis on discussing major applications of beyond graphene 2D materials in tunable devices in the past few years and an outlook of what is to come in the next decade.

cond-mat.mes-hall↗

arXiv2020

Extracting dispersion curves from ambient noise correlations using deep learning

We present a machine-learning approach to classifying the phases of surface wave dispersion curves. Standard FTAN analysis of surfaces observed on an array of receivers is converted to an image, of which, each pixel is classified as fundamental mode, first overtone, or noise. We use a convolutional neural network (U-net) architecture with a supervised learning objective and incorporate transfer learning. The training is initially performed with synthetic data to learn coarse structure, followed by fine-tuning of the network using approximately 10% of the real data based on human classification. The results show that the machine classification is nearly identical to the human picked phases. Expanding the method to process multiple images at once did not improve the performance. The developed technique will faciliate automated processing of large dispersion curve datasets.

cs.LG↗

arXiv2019

Epitaxial Growth of 2D Layered Transition Metal Dichalcogenides

Transition metal dichalcogenide (TMD) monolayers and heterostructures have emerged as a compelling class of materials with transformative new science that may be harnessed for novel device technologies. These materials are commonly fabricated by exfoliation of flakes from bulk crystals, but wafer-scale epitaxy of single crystal films is required to advance the field. This article reviews the fundamental aspects of epitaxial growth of van der Waals bonded crystals specific to TMD films. The structural and electronic properties of TMD crystals are initially described along with sources and methods used for vapor phase deposition. Issues specific to TMD epitaxy are critically reviewed including substrate properties and film-substrate orientation and bonding. The current status of TMD epitaxy on different substrate types is discussed along with characterization techniques for large area epitaxial films. Future directions are proposed including developments in substrates, in situ and full wafer characterization techniques and heterostructure growth.

cond-mat.mtrl-sci↗

arXiv2018

Dynamics of liquid crystal on hexagonal lattice

Nematic liquid crystal (LC) molecules adsorbed on two dimensional materials are aligned along the crystal directions of the hexagonal lattice. It was demonstrated that short electric pulses can reorient the aligned LC molecules in the preferred armchair direction of hexagonal boron nitride (h-BN). Several states with a variety of colors were obtained by changing the direction and strength of the electric pulses. The ab initio calculations based on density functional theory was carried out to determine the favorable adsorption configurations of the LC molecules on the h-BN surface. A non-volatile display, in which pixel resolution can be determined by grains of hexagonal surface, is proposed, which can offer a pathway towards dynamic high-quality pixels with low power consumption, and could define a new paradigm for all non-volatile display applications.

cond-mat.soft↗

arXiv2017

Efficient Determination of Reverberation Chamber Time Constant

Determination of the rate of energy loss in a reverberation chamber is fundamental to many different measurements such as absorption cross-section, antenna efficiency, radiated power, and shielding effectiveness. Determination of the energy decay time-constant in the time domain by linear fitting the power delay profile, rather than using the frequency domain quality-factor, has the advantage of being independent of the radiation efficiency of antennas used in the measurement. However, determination of chamber time constant by linear regression suffers from several practical problems, including a requirement for long measurement times. Here we present a new nonlinear curve fitting technique that can extract the time-constant with typically 60% fewer samples of the chamber transfer function for the same measurement uncertainty, which enables faster measurement of chamber time constant by sampling fewer chamber transfer function, and allows for more robust automated data post-processing. Nonlinear curve fitting could have economic benefits for test-houses, and also enables accurate broadband measurements on humans in about ten minutes for microwave exposure and medical applications. The accuracy of the nonlinear method is demonstrated by measuring the absorption cross-section of several test objects of known properties. The measurement uncertainty of the method is verified using Monte-Carlo methods.

physics.class-ph↗

arXiv2017

Mechanical Spectroscopy of Parametric Amplification in a High-Q Membrane Microresonator

We develop a stoichiometric SiN membrane-based optoelectromechanical system, in which the properties of the mechanical resonator can be dynamically controlled by the piezoelectric actuation. The mode splitting is observed in the mechanical response spectra of a phase-sensitive degenerate parametric amplifier. In addition, the quality factor Q of the membrane oscillator in this process is signicantly enhanced by more than two orders of magnitude due to the coherent amplication, reaching a Q factor of ~ 3*10^8 and a f*Q product of ~ 8*10^14 at room temperature. The thermomechanical noise squeezing close to -3 dB limitation is also achieved. This system integrates the electrical, optical and mechanical degrees of freedom without compromising the exceptional material properties of SiN membranes, and can be a useful platform for studying cavity optoelectromechanics.

physics.app-ph↗

arXiv2016

Nematic Quantum Critical Fluctuations in BaFe$_{2-x}$Ni$_x$As$_2$

We have systematically studied the nematic fluctuations in the electron-doped iron-based superconductor BaFe$_{2-x}$Ni$_x$As$_2$ by measuring the in-plane resistance change under uniaxial pressure. While the nematic quantum critical point can be identified through the measurements along the (110) direction as studied previously, quantum and thermal critical fluctuations cannot be distinguished due to similar Curie-Weiss-like behaviors. Here we find that a sizable pressure-dependent resistivity along the (100) direction is present in all doping levels, which is against the simple picture of an Ising-type nematic model. The signal along the (100) direction becomes maximum at optimal doping, suggesting that it is associated with nematic quantum critical fluctuations. Our results indicate that thermal fluctuations from striped antiferromagnetic order dominate the underdoped regime along the (110) direction. We argue that either there is a strong coupling between the quantum critical fluctuations and the fermions, or more exotically, a higher symmetry may be present around optimal doping.

cond-mat.supr-con↗

arXiv2014

Integrative Semantic Dependency Parsing via Efficient Large-scale Feature Selection

Semantic parsing, i.e., the automatic derivation of meaning representation such as an instantiated predicate-argument structure for a sentence, plays a critical role in deep processing of natural language. Unlike all other top systems of semantic dependency parsing that have to rely on a pipeline framework to chain up a series of submodels each specialized for a specific subtask, the one presented in this article integrates everything into one model, in hopes of achieving desirable integrity and practicality for real applications while maintaining a competitive performance. This integrative approach tackles semantic parsing as a word pair classification problem using a maximum entropy classifier. We leverage adaptive pruning of argument candidates and large-scale feature selection engineering to allow the largest feature space ever in use so far in this field, it achieves a state-of-the-art performance on the evaluation data set for CoNLL-2008 shared task, on top of all but one top pipeline system, confirming its feasibility and effectiveness.

cs.CL↗

arXiv2013

Loop Current and Orbital Magnetism in Pyrite-Type $Ir_{3}Ch_{8}$ ($Ch=Se,Te$)

We study the parent compound of the pyrite-type $Ir_{x}Ch_{2}$ iridium chalcogenide superconductors, $Ir_{0.75}Ch_{2}$ (or $Ir_{3}Ch_{8}$). While the lattice structure of the material is rather complicated, we show that the electron physics near the Fermi surface (FS) can be described by a three-band tight-binding model. We find that the material possesses a loop current ground state which is responsible for the highly anisotropic strong diamagnetism observed in experiments. The fluctuations of this loop current state can be a possible pairing mechanism for superconductivity.

cond-mat.supr-con↗

arXiv2013

A magnetic origin for high temperature superconductivity in iron pnictides

In conventional Bardeen-Cooper-Schrieffer (BCS) superconductors, superconductivity occurs when electrons form coherent Cooper pairs below the superconducting transition temperature Tc. Although the kinetic energy of paired electrons increases in the superconducting state relative to the normal state, the reduction in the ion lattice energy is sufficient to give the superconducting condensation energy. For iron pnictide superconductors derived from electron or hole doping of their antiferromagnetic (AF) parent compounds, the microscopic origin for supercnductivity is unclear. Here we use neutron scattering to show that high-Tc superconductivity only occurs for iron pnictides with low-energy itinerant electron-spin excitation coupling and high energy spin excitations. Since our absolute spin susceptibility measurements for optimally hole-doped iron pnictide reveal that the change in magnetic exchange energy below and above Tc can account for the superconducting condensation energy, we conclude that the presence of both high-energy spin excitations giving rise to a large magnetic exchange coupling J and low-energy spin excitations coupled to the itinerant electrons is essential for high-Tc superconductivity in iron pnictides.

cond-mat.supr-con↗

arXiv2012

Strong-Coupling Superconductivity in NaFe$_{1-x}$Co$_x$As: the Eliashberg Theory and Beyond

We study the normal-state and superconducting properties of NaFe$_{1-x}$Co$_x$As system by specific heat measurements. Both the normal-state Sommerfeld coefficient and superconducting condensation energy are strongly suppressed in the underdoped and heavily overdoped samples. The low-temperature electronic specific heat can be well fitted by either an one-gap or a two-gap BCS-type function for all the superconducting samples. The ratio $γ_NT_c^2/H_c^2(0)$ can nicely associate the neutron spin resonance as the bosons in the standard Eliashberg model. However, the value of $ΔC/T_cγ_N$ near optimal doping is larger than the maximum value the model can obtain. Our results suggest that the high-$T_c$ superconductivity in the Fe-based superconductors may be understood within the framework of boson-exchange mechanism but significant modification may be needed to account for the finite-temperature properties.

cond-mat.supr-con↗

arXiv2012

Effect of Li-deficiency impurities on the electron-overdoped LiFeAs superconductor

We use transport, inelastic neutron scattering, and angle resolved photoemission experiments to demonstrate that the stoichiometric LiFeAs is an intrinsically electron-overdoped superconductor similar to those of the electron-overdoped NaFe1-xTxAs and BaFe2-xTxAs2 (T = Co,Ni). Furthermore, we show that although transport properties of the stoichiometric superconducting LiFeAs and Li-deficient nonsuperconducting Li1-xFeAs are different, their electronic and magnetic properties are rather similar. Therefore, the nonsuperconducting Li1-xFeAs is also in the electron overdoped regime, where small Li deficiencies near the FeAs octahedra can dramatically suppress superconductivity through the impurity scattering effect.

cond-mat.supr-con↗

arXiv2011

Microscopic model for the ferroelectric field effect in oxide heterostructures

A microscopic model Hamiltonian for the ferroelectric field effect is introduced for the study of oxide heterostructures with ferroelectric components. The long-range Coulomb interaction is incorporated as an electrostatic potential, solved self-consistently together with the charge distribution. A generic double-exchange system is used as the conducting channel, epitaxially attached to the ferroelectric gate. The observed ferroelectric screening effect, namely the charge accumulation/depletion near the interface, is shown to drive interfacial phase transitions that give rise to robust magnetoelectric responses and bipolar resistive switching, in qualitative agreement with previous density functional theory calculations. The model can be easily adapted to other materials by modifying the Hamiltonian of the conducting channel, and it is useful in simulating ferroelectric field effect devices particularly those involving strongly correlated electronic components where ab-initio techniques are difficult to apply.

cond-mat.mtrl-sci↗

arXiv2011

Optical Conductivity Anisotropy in the Undoped Three-Orbital Hubbard Model for the Pnictides

The resistivity anisotropy unveiled in the study of detwinned single crystals of the undoped 122 pnictides is here studied using the two-dimensional three-orbital Hubbard model in the mean-field approximation. Calculating the Drude weight in the x and y directions at zero temperature for a Q=(π,0) magnetically ordered state, the conductance along the antiferromagnetic direction is shown to be larger than along the ferromagnetic direction. This effect is caused by the suppression of the d_{yz} orbital at the Fermi surface, but additional insight based on the momentum dependence of the transitions induced by the current operator is provided. It is shown that the effective suppression of the inter-orbital hopping d_{xy} and d_{yz} along the y direction is the main cause of the anisotropy.

cond-mat.supr-con↗