arXiv subjects

Yu Deng

Publications and source records attributed to Yu Deng.

At least 109 records · Page 6Linked to original sources

arXiv2021

BERTSurv: BERT-Based Survival Models for Predicting Outcomes of Trauma Patients

Survival analysis is a technique to predict the times of specific outcomes, and is widely used in predicting the outcomes for intensive care unit (ICU) trauma patients. Recently, deep learning models have drawn increasing attention in healthcare. However, there is a lack of deep learning methods that can model the relationship between measurements, clinical notes and mortality outcomes. In this paper we introduce BERTSurv, a deep learning survival framework which applies Bidirectional Encoder Representations from Transformers (BERT) as a language representation model on unstructured clinical notes, for mortality prediction and survival analysis. We also incorporate clinical measurements in BERTSurv. With binary cross-entropy (BCE) loss, BERTSurv can predict mortality as a binary outcome (mortality prediction). With partial log-likelihood (PLL) loss, BERTSurv predicts the probability of mortality as a time-to-event outcome (survival analysis). We apply BERTSurv on Medical Information Mart for Intensive Care III (MIMIC III) trauma patient data. For mortality prediction, BERTSurv obtained an area under the curve of receiver operating characteristic curve (AUC-ROC) of 0.86, which is an improvement of 3.6% over baseline of multilayer perceptron (MLP) without notes. For survival analysis, BERTSurv achieved a concordance index (C-index) of 0.7. In addition, visualizations of BERT's attention heads help to extract patterns in clinical notes and improve model interpretability by showing how the model assigns weights to different inputs.

cs.AI↗

arXiv2021

1550 nm compatible ultrafast photoconductive material based on a GaAs/ErAs/GaAs heterostructure

The sub-bandgap absorption and ultrafast relaxation in a GaAs/ErAs/GaAs heterostructure are reported. The infrared absorption and 1550 nm-excited ultrafast photo-response are studied by Fourier transform infrared (FTIR) spectrometry and time-domain pump-probe technique. The two absorption peaks located at 2.0 um (0.62 eV) and 2.7 um (0.45 eV) are originated from the ErAs/GaAs interfacial Schottky states and sub-bandgap transition within GaAs, respectively. The photo-induced carrier lifetime, excited using 1550 nm light, is measured to be as low as 190 fs for the GaAs/ErAs/GaAs heterostructure, making it a promising material for 1550-nm-technology-compatible, high critical-breakdown-field THz devices. The relaxation mechanism is proposed and the functionality of ErAs is revealed.

cond-mat.mtrl-sci↗

arXiv2021

A Deep Learning-Based Approach to Extracting Periosteal and Endosteal Contours of Proximal Femur in Quantitative CT Images

Automatic CT segmentation of proximal femur is crucial for the diagnosis and risk stratification of orthopedic diseases; however, current methods for the femur CT segmentation mainly rely on manual interactive segmentation, which is time-consuming and has limitations in both accuracy and reproducibility. In this study, we proposed an approach based on deep learning for the automatic extraction of the periosteal and endosteal contours of proximal femur in order to differentiate cortical and trabecular bone compartments. A three-dimensional (3D) end-to-end fully convolutional neural network, which can better combine the information between neighbor slices and get more accurate segmentation results, was developed for our segmentation task. 100 subjects aged from 50 to 87 years with 24,399 slices of proximal femur CT images were enrolled in this study. The separation of cortical and trabecular bone derived from the QCT software MIAF-Femur was used as the segmentation reference. We randomly divided the whole dataset into a training set with 85 subjects for 10-fold cross-validation and a test set with 15 subjects for evaluating the performance of models. Two models with the same network structures were trained and they achieved a dice similarity coefficient (DSC) of 97.87% and 96.49% for the periosteal and endosteal contours, respectively. To verify the excellent performance of our model for femoral segmentation, we measured the volume of different parts of the femur and compared it with the ground truth and the relative errors between predicted result and ground truth are all less than 5%. It demonstrated a strong potential for clinical use, including the hip fracture risk prediction and finite element analysis.

cs.CV↗

arXiv2021

Invariant Gibbs measure and global strong solutions for the Hartree NLS equation in dimension three

In this paper we consider the defocusing Hartree nonlinear Schrödinger equations on $\mathbb T^3$ with real valued and even potential $V$ and Fourier multiplier decaying like $|k|^{-β}$. By relying on the method of random averaging operators in arXiv:1910.08492, we show that there exists $\frac{1}{2} \ll β_0 <1 $ such that for $ β> β_0 $ we have invariance of the associated Gibbs measure and global existence of strong solutions in its statistical ensemble. In this way we extend Bourgain's seminal result [7] which requires $β>2$ in this case.

math.AP↗

arXiv2021

Spread Mechanism and Influence Measurement of Online Rumors in China During the COVID-19 Pandemic

In early 2020, the Corona Virus Disease 2019 (COVID-19) pandemic swept the world.In China, COVID-19 has caused severe consequences. Moreover, online rumors during the COVID-19 pandemic increased people's panic about public health and social stability. At present, understanding and curbing the spread of online rumors is an urgent task. Therefore, we analyzed the rumor spreading mechanism and propose a method to quantify a rumors' influence by the speed of new insiders. The search frequency of the rumor is used as an observation variable of new insiders. The peak coefficient and the attenuation coefficient are calculated for the search frequency, which conforms to the exponential distribution. We designed several rumor features and used the above two coefficients as predictable labels. A 5-fold cross-validation experiment using the mean square error (MSE) as the loss function showed that the decision tree was suitable for predicting the peak coefficient, and the linear regression model was ideal for predicting the attenuation coefficient. Our feature analysis showed that precursor features were the most important for the outbreak coefficient, while location information and rumor entity information were the most important for the attenuation coefficient. Meanwhile, features that were conducive to the outbreak were usually harmful to the continued spread of rumors. At the same time, anxiety was a crucial rumor causing factor. Finally, we discuss how to use deep learning technology to reduce the forecast loss by using the Bidirectional Encoder Representations from Transformers (BERT) model.

cs.SI↗

arXiv2020

Large magnetoresistance observed in α-Sn/InSb heterostructures

In this study, we report the epitaxial growth of a series of α-Sn films on InSb substrate by molecular beam epitaxy (MBE) with thickness varying from 10 nm to 400 nm. High qualities of the α-Sn films are confirmed. An enhanced large magnetoresistance (MR) over 450,000% has been observed compared to that of the bare InSb substrate. Thickness, angle and temperature dependent MR are used to demonstrate the effects of α-Sn films on the electrical transport properties.

cond-mat.mtrl-sci↗

arXiv2020

Disentangled and Controllable Face Image Generation via 3D Imitative-Contrastive Learning

We propose DiscoFaceGAN, an approach for face image generation of virtual people with disentangled, precisely-controllable latent representations for identity of non-existing people, expression, pose, and illumination. We embed 3D priors into adversarial learning and train the network to imitate the image formation of an analytic 3D face deformation and rendering process. To deal with the generation freedom induced by the domain gap between real and rendered faces, we further introduce contrastive learning to promote disentanglement by comparing pairs of generated images. Experiments show that through our imitative-contrastive learning, the factor variations are very well disentangled and the properties of a generated face can be precisely controlled. We also analyze the learned latent space and present several meaningful properties supporting factor disentanglement. Our method can also be used to embed real images into the disentangled latent space. We hope our method could provide new understandings of the relationship between physical properties and deep image synthesis.

cs.CV↗

arXiv2020

Crossing Variational Autoencoders for Answer Retrieval

Answer retrieval is to find the most aligned answer from a large set of candidates given a question. Learning vector representations of questions/answers is the key factor. Question-answer alignment and question/answer semantics are two important signals for learning the representations. Existing methods learned semantic representations with dual encoders or dual variational auto-encoders. The semantic information was learned from language models or question-to-question (answer-to-answer) generative processes. However, the alignment and semantics were too separate to capture the aligned semantics between question and answer. In this work, we propose to cross variational auto-encoders by generating questions with aligned answers and generating answers with aligned questions. Experiments show that our method outperforms the state-of-the-art answer retrieval method on SQuAD.

cs.CL↗

arXiv2020

Random tensors, propagation of randomness, and nonlinear dispersive equations

Abstract. The purpose of this paper is twofold. We introduce the theory of random tensors, which naturally extends the method of random averaging operators in our earlier work arXiv:1910.08492, to study the propagation of randomness under nonlinear dispersive equations. By applying this theory we also solve Conjecture 1.7 in arXiv:1910.08492, and establish almost-sure local well-posedness for semilinear Schrödinger equations in spaces that are subcritical in the probabilistic scaling. The solution we find has an explicit expansion in terms of multilinear Gaussians with adapted random tensor coefficients. In the random setting, the probabilistic scaling is the natural scaling for dispersive equations, and is different from the natural scaling for parabolic equations. Our theory, which covers the full subcritical regime in the probabilistic scaling, can be viewed as the dispersive counterpart of the existing parabolic theories (regularity structure, para-controlled calculus and renormalization group techniques).

math.AP↗

arXiv2020

Deep 3D Portrait from a Single Image

In this paper, we present a learning-based approach for recovering the 3D geometry of human head from a single portrait image. Our method is learned in an unsupervised manner without any ground-truth 3D data. We represent the head geometry with a parametric 3D face model together with a depth map for other head regions including hair and ear. A two-step geometry learning scheme is proposed to learn 3D head reconstruction from in-the-wild face images, where we first learn face shape on single images using self-reconstruction and then learn hair and ear geometry using pairs of images in a stereo-matching fashion. The second step is based on the output of the first to not only improve the accuracy but also ensure the consistency of overall head geometry. We evaluate the accuracy of our method both in 3D and with pose manipulation tasks on 2D images. We alter pose based on the recovered geometry and apply a refinement network trained with adversarial learning to ameliorate the reprojected images and translate them to the real image domain. Extensive evaluations and comparison with previous methods show that our new method can produce high-fidelity 3D head geometry and head pose manipulation results.

cs.CV↗

arXiv2020

Accurate 3D Face Reconstruction with Weakly-Supervised Learning: From Single Image to Image Set

Recently, deep learning based 3D face reconstruction methods have shown promising results in both quality and efficiency.However, training deep neural networks typically requires a large volume of data, whereas face images with ground-truth 3D face shapes are scarce. In this paper, we propose a novel deep 3D face reconstruction approach that 1) leverages a robust, hybrid loss function for weakly-supervised learning which takes into account both low-level and perception-level information for supervision, and 2) performs multi-image face reconstruction by exploiting complementary information from different images for shape aggregation. Our method is fast, accurate, and robust to occlusion and large pose. We provide comprehensive experiments on three datasets, systematically comparing our method with fifteen recent methods and demonstrating its state-of-the-art performance.

cs.CV↗

arXiv2019

On the derivation of the wave kinetic equation for NLS

A fundamental question in wave turbulence theory is to understand how the "wave kinetic equation" (WKE) describes the long-time dynamics of its associated nonlinear dispersive equation. Formal derivations in the physics literature date back to the work of Pieirls in 1928. For the cubic nonlinear Schrödinger equation, it is expected that such a kinetic description should hold, in a limiting regime where the size $L$ of the domain goes to infinity, and the strength $α$ of the nonlinearity goes to 0 (weak nonlinearity), at a kinetic time scale $T_{\mathrm{kin}}=O(α^{-2})$. In this paper, we study the rigorous justification of this monumental statement, and show that the answer seems to depend on the particular "scaling law" in which the $(α, L)$ limit is taken, in a spirit similar to how the Boltzmann-Grad scaling law is imposed in the derivation of Boltzmann's equation. In particular, there appears to be two favorable scaling laws: when $α$ approaches $0$ like $L^{-\varepsilon+}$ or like $L^{-1-\frac{\varepsilon}{2}+}$ (for arbitrary small $\varepsilon$), we exhibit the wave kinetic equation up to timescales $O(T_{\mathrm{kin}}L^{-\varepsilon})$, by showing that the relevant Feynman diagram expansions converge absolutely (as a sum over paired trees). For the other scaling laws, we justify the onset of the kinetic description at timescales $T_*\ll T_{\mathrm{kin}}$, and identify specific interactions that become very large for times beyond $T_*$. In particular, the relevant tree expansion diverges absolutely there. In light of those interactions, extending the kinetic description beyond $T_*$ towards $T_{\mathrm{kin}}$ for such scaling laws seems to require new methods and ideas.

math.AP↗

arXiv2019

On the Smallness Condition in Linear Inviscid Damping: Monotonicity and Resonance Chains

We consider the linearized Euler equations around a smooth, bilipschitz shear profile $U(y)$ on $\mathbb{T}_L \times \mathbb{R}$. We construct an explicit flow which exhibits linear inviscid damping for $L$ sufficiently small, but for which damping fails if $L$ is large. In particular, similar to the instability results for convex profiles for a shear flow being bilipschitz is not sufficient for linear inviscid damping to hold. Instead of an eigenvalue-based argument the underlying mechanism here is shown to be based on a new cascade of resonances moving to higher and higher frequencies in $y$, which is distinct from the echo chain mechanism in the nonlinear problem.

math.AP↗

arXiv2019

Automatic phase space generation for Monte Carlo calculations of Intensity Modulated Particle Therapy

Monte Carlo (MC) is generally considered as the most accurate dose calculation tool for particle therapy. However, a proper description of the beam particles kinematics is a necessary input for a realistic simulation. Such a description can be stored in phase space (PS) files for different beam energies. A PS file contains kinetic information such as energies, positions and travelling directions for particles traversing a plane perpendicular to the beam direction. The accuracy of PS files play a critical role in the performance of the MC method for dose calculations. A PS file can be generated with a set of parameters describing analytically the beam kinematics. However, determining such parameters can be tedious and time consuming. Thus, we have developed an algorithm to obtain those parameters automatically and efficiently. In this paper, we present such an algorithm and compare dose calculations using PS automatically generated for the Shanghai Proton and Heavy Ion Center (SPHIC) with measurements. The gamma-index for comparing calculated depth dose distributions (DDD) with measurements are above 96.0% with criterion 0.6%/0.6 mm. For each single energy, the mean difference percentage between calculated lateral spot sizes at 5 different depths and measurements are below 3.5%.

physics.med-ph↗

arXiv2019

Echo Chains as a Linear Mechanism: Norm Inflation, Modified Exponents and Asymptotics

In this article we show that the Euler equations, when linearized around a low frequency perturbation to Couette flow, exhibit norm inflation in Gevrey-type spaces as time tends to infinity. Thus, echo chains are shown to be a (secondary) linear instability mechanism. Furthermore, we develop a more precise analysis of cancellations in the resonance mechanism, which yields a modified exponent in the high frequency regime. In addition it allows us to remove a logarithmic constraint on the perturbations present in prior works by Bedrossian, Deng and Masmoudi and to construct solutions which are initially in a Gevrey class for which the velocity asymptotically converges in Sobolev regularity but diverges in Gevrey regularity.

math.AP↗

arXiv2019

Photoluminescence mapping and time-domain thermo-photoluminescence for rapid imaging and measurement of thermal conductivity of boron arsenide

Cubic boron arsenide (BAs) is attracting greater attention due to the recent experimental demonstration of ultrahigh thermal conductivity \k{appa} above 1000 W/mK. However, its bandgap has not been settled and a simple yet effective method to probe its crystal quality is missing. Furthermore, traditional \k{appa} measurement methods are destructive and time consuming, thus they cannot meet the urgent demand for fast screening of high \k{appa} materials. After we experimentally established 1.82 eV as the indirect bandgap of BAs and observed room-temperature band-edge photoluminescence, we developed two new optical techniques that can provide rapid and non-destructive characterization of \k{appa} with little sample preparation: photoluminescence mapping (PL-mapping) and time-domain thermo-photoluminescence (TDTP). PL-mapping provides nearly real-time image of crystal quality and \k{appa} over mm-sized crystal surfaces; while TDTP allows us to pick up any spot on the sample surface and measure its \k{appa} using nanosecond laser pulses. These new techniques reveal that the apparent single crystals are not only non-uniform in \k{appa}, but also are made of domains of very distinct \k{appa}. Because PL-mapping and TDTP are based on the band-edge PL and its dependence on temperature, they can be applied to other semiconductors, thus paving the way for rapid identification and development of high-\k{appa} semiconducting materials.

cond-mat.mtrl-sci↗

arXiv2019

Thermal Stability Enhancement in Epitaxial α-Sn Films by Strain Engineering

Exploring new topological materials with large topological nontrivial bandgaps and simple composition is attractive for both theoretical investigation and experimental realization. Recently alpha tin (α-Sn) has been predicted to be such a candidate and it can be tuned to be either a topological insulator or a Dirac semimetal by applying appropriate strain. However, free-standing α-Sn is only stable below 13.2 °C. In this study, a series of high-quality α-Sn films with different thicknesses have been successfully grown on InSb substrates by molecular beam epitaxy (MBE). Confirmed by both X-ray diffraction (XRD) and reciprocal space mapping (RSM), all the films remained fully strained up to 4000 Å, proving the strain effect from the substrate. Remarkably, the single-crystalline α phase can persist up to 170 °C for the 200 Å thick sample. The critical temperature where the α phase disappears decreases as the film thickness increases, showing the thermal stabilization can be engineered by varying the α-Sn thickness. A plastic flow model taking the work hardening into account is introduced to explain this dependence, assuming the strain relaxation and the phase transition occur successively. This enhanced thermal stability is prerequisite for above room-temperature characterization and application of this material system.

cond-mat.mtrl-sci↗

arXiv2019

Optimal local well-posedness for the periodic derivative nonlinear Schrodinger equation

We prove local well-posedness for the periodic derivative nonlinear Schrodinger's equation, which is L^2 critical, in Fourier-Lebesgue spaces which scale like H^s(T) for s>0. In particular we close the existing gap in the subcritical theory by improving the result of Grunrock and Herr [25], which established local well-posedness in Fourier-Lebesgue spaces which scale like H^s(T) for s>1 . We achieve this result by a delicate analysis of the structure of the solution and the construction of an adapted nonlinear submanifold of a suitable function space. Together these allow us to construct the unique solution to the given subcritical data. This constructive procedure is inspired by the theory of para-controlled distributions developed by Gubinelli-Imkeller-Perkowski [26] and Cantellier-Chouk [10] in the context of stochastic PDE. Our proof and results however, are purely deterministic.

math.AP↗