arXiv subjects

Hui Xie

Publications and source records attributed to Hui Xie.

28 records · Page 2Linked to original sources

arXiv2022

A weighted first-order formulation for solving anisotropic diffusion equations with deep neural networks

In this paper, a new weighted first-order formulation is proposed for solving the anisotropic diffusion equations with deep neural networks. For many numerical schemes, the accurate approximation of anisotropic heat flux is crucial for the overall accuracy. In this work, the heat flux is firstly decomposed into two components along the two eigenvectors of the diffusion tensor, thus the anisotropic heat flux approximation is converted into the approximation of two isotropic components. Moreover, to handle the possible jump of the diffusion tensor across the interface, the weighted first-order formulation is obtained by multiplying this first-order formulation by a weighted function. By the decaying property of the weighted function, the weighted first-order formulation is always well-defined in the pointwise way. Finally, the weighted first-order formulation is solved with deep neural network approximation. Compared to the neural network approximation with the original second-order elliptic formulation, the proposed method can significantly improve the accuracy, especially for the discontinuous anisotropic diffusion problems.

math.NA↗

arXiv2021

Dual camera snapshot hyperspectral imaging system via physics informed learning

We consider using the system's optical imaging process with convolutional neural networks (CNNs) to solve the snapshot hyperspectral imaging reconstruction problem, which uses a dual-camera system to capture the three-dimensional hyperspectral images (HSIs) in a compressed way. Various methods using CNNs have been developed in recent years to reconstruct HSIs, but most of the supervised deep learning methods aimed to fit a brute-force mapping relationship between the captured compressed image and standard HSIs. Thus, the learned mapping would be invalid when the observation data deviate from the training data. Especially, we usually don't have ground truth in real-life scenarios. In this paper, we present a self-supervised dual-camera equipment with an untrained physics-informed CNNs framework. Extensive simulation and experimental results show that our method without training can be adapted to a wide imaging environment with good performance. Furthermore, compared with the training-based methods, our system can be constantly fine-tuned and self-improved in real-life scenarios.

eess.IV↗

arXiv2020

Globally Optimal Segmentation of Mutually Interacting Surfaces using Deep Learning

Segmentation of multiple surfaces in medical images is a challenging problem, further complicated by the frequent presence of weak boundary and mutual influence between adjacent objects. The traditional graph-based optimal surface segmentation method has proven its effectiveness with its ability of capturing various surface priors in a uniform graph model. However, its efficacy heavily relies on handcrafted features that are used to define the surface cost for the "goodness" of a surface. Recently, deep learning (DL) is emerging as powerful tools for medical image segmentation thanks to its superior feature learning capability. Unfortunately, due to the scarcity of training data in medical imaging, it is nontrivial for DL networks to implicitly learn the global structure of the target surfaces, including surface interactions. In this work, we propose to parameterize the surface cost functions in the graph model and leverage DL to learn those parameters. The multiple optimal surfaces are then simultaneously detected by minimizing the total surface cost while explicitly enforcing the mutual surface interaction constraints. The optimization problem is solved by the primal-dual Internal Point Method, which can be implemented by a layer of neural networks, enabling efficient end-to-end training of the whole network. Experiments on Spectral Domain Optical Coherence Tomography (SD-OCT) retinal layer segmentation and Intravascular Ultrasound (IVUS) vessel wall segmentation demonstrated very promising results. All source code is public to facilitate further research at this direction.

cs.CV↗

arXiv2020

Hydrogen "penta-graphene-like" structure stabilized by hafnium: a high-temperature conventional superconductor

The recent discovery of H3S and LaH10 superconductors with record high superconducting transition temperatures, Tc, at high pressure, has fueled the search for room-temperature superconductivity in the compressed superhydrides. Here we predict the existence of an unprecedented hexagonal HfH10, with an extraordinarily high Tc of around 213-234 K at 250 GPa. In HfH10, the H atoms are arranged in clusters to form a planar "penta-graphene-like" sublattice, in contrast to the covalent sixfold cubic structure in H3S and clathrate-like structure in LaH10. The Hf atom acts as a "precompressor" and electron donor to the hydrogen sublattice. This "penta-graphene-like" H10 structure is also found in ZrH10, ScH10 and LuH10 at high pressure, each material showing a high Tc ranging from 134 to 220 kelvin. Our study of dense superhydrides with "penta-graphene-like" layered structures opens the door to the exploration and exploitation of a new class of high Tc superconductors.

cond-mat.supr-con↗

arXiv2020

High-Pressure Synthesis of Magnetic Neodymium Polyhydrides

The current search for room-temperature superconductivity is inspired by the unique properties of the electron-phonon interaction in metal superhydrides. Encouraged by the recently found highest-$T_C$ superconductor fcc-$LaH_{10}$, here we discover several superhydrides of another lanthanide - neodymium. We identify three novel metallic Nd-H phases at pressure range from 85 to 135 GPa: $I4/mmm$-$NdH_4$, $C2/c$-$NdH_7$, $P6_3/mmc$-$NdH_9$, synthesized by laser-heating metal samples in NH3BH3 media for in situ generation of hydrogen. A lower trihydride $Fm\bar{3}m$-$NdH_3$ is found at pressures from 2 to 52 GPa. $I4/mmm$-$NdH_4$ and $C2/c$-$NdH_7$ are stable from 135 down to 85 GPa, and $P6_3/mmc$-$NdH_9$ from 110 to 130 GPa. Measurements of the electrical resistance of NdH9 demonstrate a possible superconducting transition at ~4.5 K in $P6_3/mmc$-$NdH_9$. Our theoretical calculations predict that all the neodymium hydrides have antiferromagnetic order at pressures below 150 GPa and represent one of the first discovered examples of strongly correlated superhydrides with large exchange spin-splitting in the electron band structure (> 450 meV). The critical N$é$el temperatures for new neodymium hydrides are estimated using the mean-field approximation as about 4 K ($NdH_4$), 251 K ($NdH_7$) and 136 K ($NdH_9$).

cond-mat.supr-con↗

arXiv2019

Superconducting Praseodymium Superhydrides

Superhydrides have complex hydrogenic sublattices and are important prototypes for studying metallic hydrogen and high-temperature superconductors. Encouraged by the results on LaH10, in consideration of the differences between La and Pr, Pr-H system is especially worth studying because of the magnetism and valence-band f-electrons in element Pr. Here we successfully synthesized praseodymium superhydrides (PrH9) in laser-heated diamond anvil cells. Synchrotron X-ray diffraction (XRD) analysis demonstrated the presence of previously predicted F43m-PrH9 and unexpected P63/mmc-PrH9 phases. Moreover, Fm3m-PrH3, P4/nmm-PrH(3-δ) and Fm3m-PrH(1+x) were found below 52 GPa. F43m-PrH9 and P63/mmc-PrH9 were stable above 100 GPa in experiment. Experimental studies of electrical resistance in the PrH9 sample showed the emergence of superconducting transition (Tc) below 9K and a dependent Tc on applied magnetic field. Theoretical calculations indicate that magnetic order and electron-phonon interaction coexist in a very close range of pressures in the PrH9 sample which may contribute to its low superconducting temperature Tc. Our results highlight the intimate connections among hydrogenic sublattices, density of states, magnetism and superconductivity in Pr-based superhydrides.

cond-mat.mtrl-sci↗

arXiv2019

Trust but Verify: An Information-Theoretic Explanation for the Adversarial Fragility of Machine Learning Systems, and a General Defense against Adversarial Attacks

Deep-learning based classification algorithms have been shown to be susceptible to adversarial attacks: minor changes to the input of classifiers can dramatically change their outputs, while being imperceptible to humans. In this paper, we present a simple hypothesis about a feature compression property of artificial intelligence (AI) classifiers and present theoretical arguments to show that this hypothesis successfully accounts for the observed fragility of AI classifiers to small adversarial perturbations. Drawing on ideas from information and coding theory, we propose a general class of defenses for detecting classifier errors caused by abnormally small input perturbations. We further show theoretical guarantees for the performance of this detection method. We present experimental results with (a) a voice recognition system, and (b) a digit recognition system using the MNIST database, to demonstrate the effectiveness of the proposed defense methods. The ideas in this paper are motivated by a simple analogy between AI classifiers and the standard Shannon model of a communication system.

cs.CR↗

arXiv2019

Photoelectron holographic interferometry to probe the longitudinal momentum offset at the tunnel exit

Laser-induced electron tunneling underlies numerous emerging spectroscopic techniques to probe attosecond electron dynamics in atoms and molecules. The improvement of those techniques requires an accurate knowledge of the exit momentum for the tunneling wave packet. Here we demonstrate a photoelectron interferometric scheme to probe the electron momentum longitudinal to the tunnel direction at the tunnel exit by measuring the photoelectron holographic pattern in an orthogonally polarized two-color laser pulse. In this scheme, we use a perturbative 400-nm laser field to modulate the photoelectron holographic fringes generated by a strong 800-nm pulse. The fringe shift offers a direct experimental access to the intermediate canonical momentum of the rescattering electron, allowing us to reconstruct the momentum offset at the tunnel exit with high accuracy. Our result unambiguously proves the existence of nonzero initial longitudinal momentum at the tunnel exit and provides fundamental insights into the non-quasi-static nature of the strong-field tunneling.

physics.atom-ph↗

arXiv2019

An Information-Theoretic Explanation for the Adversarial Fragility of AI Classifiers

We present a simple hypothesis about a compression property of artificial intelligence (AI) classifiers and present theoretical arguments to show that this hypothesis successfully accounts for the observed fragility of AI classifiers to small adversarial perturbations. We also propose a new method for detecting when small input perturbations cause classifier errors, and show theoretical guarantees for the performance of this detection method. We present experimental results with a voice recognition system to demonstrate this method. The ideas in this paper are motivated by a simple analogy between AI classifiers and the standard Shannon model of a communication system.

cs.IT↗

arXiv2017

The fluid running in the subnanochannel with functional surface

We have researched the motion of gas in the subnanochannel with functional surface which wettability has a gradient for the fluid by using molecular dynamics simulation. The results show that the gas is driven to flow under a single heat source and without any other work or energy applied to the system. The driving source is owed to the potential gradient of the functional face which keeps the fluid running in the subnanochannel. The width of the channel and the pressure of the reservoir has a significant influence on the flow velocity, which, respectively, has an optimal value for the maximum velocity. The flow velocity grows with the increasing temperature.

physics.comp-ph↗