arXiv subjects

Yue Hu

Publications and source records attributed to Yue Hu.

At least 289 records · Page 16Linked to original sources

arXiv2020

Modeling of Galactic Foreground Polarization with Velocity Gradients

The detection of primordial B-mode polarization is still challenging due to the relatively low amplitude compared to the galactic foregrounds. To remove the contribution from the foreground, a comprehensive picture of the galactic magnetic field is indispensable. The Velocity Gradient Technique (VGT) is promising in tracing magnetic fields based on the modern understanding of the magneto-hydrodynamic turbulence. In this work, we apply VGT to a HI region containing an intermediate velocity cloud and a local velocity cloud, which are distinguishable in position-position-velocity space. We show that VGT gives an excellent agreement with the Planck polarization and stellar polarization. We confirm the advantages of VGT in constructing the 3D galactic magnetic field.

astro-ph.GA↗

arXiv2020

IIE-NLP-NUT at SemEval-2020 Task 4: Guiding PLM with Prompt Template Reconstruction Strategy for ComVE

This paper introduces our systems for the first two subtasks of SemEval Task4: Commonsense Validation and Explanation. To clarify the intention for judgment and inject contrastive information for selection, we propose the input reconstruction strategy with prompt templates. Specifically, we formalize the subtasks into the multiple-choice question answering format and construct the input with the prompt templates, then, the final prediction of question answering is considered as the result of subtasks. Experimental results show that our approaches achieve significant performance compared with the baseline systems. Our approaches secure the third rank on both official test sets of the first two subtasks with an accuracy of 96.4 and an accuracy of 94.3 respectively.

cs.CL↗

arXiv2020

Anomalous thermal transport in metallic transition-metal nitrides originated from strong electron-phonon interactions

Metallic transition-metal nitrides (TMNs) are promising conductive ceramics for many applications, whose thermal transport is of great importance in device design. It is found metallic TiN and HfN hold anomalous thermal transport behaviors compared to common metals and nonmetallic TMNs. They have extremely large intrinsic phonon thermal conductivity mainly due to the large acoustic-optic phonon frequency gaps. The phonon thermal conductivity is reduced by two orders of magnitude as the phonon-isotope and phonon-electron scatterings are considered, which also induce the nontrivial temperature-independent behavior of phonon thermal conductivity. Nesting Fermi surfaces exist in both TiN and HfN, which cause the strong electron-phonon coupling strengths and heavily harm the transport of phonons and electrons. The phonon component takes an abnormally large ratio in total thermal conductivity, as 29% for TiN and 26% for HfN at 300 K. The results for thin films are also presented and it is shown that the phonon thermal conductivity can be efficiently limited by size. Our findings provide a deep understanding on the thermal transport in metallic TMNs and expand the scope of heat conduction theory in metal.

cond-mat.mtrl-sci↗

arXiv2020

Velocity Gradient in the Presence of Self-Gravity: Identifying Gravity-induced Inflow and Determining Collapsing Stage

Understanding how star formation is regulated requires studying the energy balance between turbulence, magnetic fields, stellar feedback, and gravity within molecular clouds. However, identifying the transition region where the gravity takes over remains elusive. Recent studies of the Velocity Gradient Technique (VGT), which is an advanced tool for magnetic field studies, reveal that the gradients of spectroscopic observables change their directions by 90 degrees with respect to the magnetic fields in the regions of gravitational collapse. In this study, we perform 3D MHD numerical simulations. We observe that star formation successfully proceeds in strongly magnetized and fully ionized media. We confirm that the self-gravity induces the change of gradients' orientation and high gradients' amplitude. We explore two ways of identifying collapsing self-gravitating regions through the double-peak feature in the histogram of gradients' orientation and the curvature of gradients. We show that velocity gradients' morphology and amplitude can be synthetically used to trace the convergent inflows. By comparing with the column density Probability Density Functions (N-PDFs) method, we show that VGT is a powerful new tool for studying the gas dynamics and tracing magnetic field in star-forming regions. By analogy with VGT, we extend the Intensity Gradient Technique (IGT) to locate the gravitational collapsing region and shocks. We demonstrate the synergy of VGT and IGT can determine the collapsing stages in a star-forming region.

astro-ph.GA↗

arXiv2020

Multiscale Collaborative Deep Models for Neural Machine Translation

Recent evidence reveals that Neural Machine Translation (NMT) models with deeper neural networks can be more effective but are difficult to train. In this paper, we present a MultiScale Collaborative (MSC) framework to ease the training of NMT models that are substantially deeper than those used previously. We explicitly boost the gradient back-propagation from top to bottom levels by introducing a block-scale collaboration mechanism into deep NMT models. Then, instead of forcing the whole encoder stack directly learns a desired representation, we let each encoder block learns a fine-grained representation and enhance it by encoding spatial dependencies using a context-scale collaboration. We provide empirical evidence showing that the MSC nets are easy to optimize and can obtain improvements of translation quality from considerably increased depth. On IWSLT translation tasks with three translation directions, our extremely deep models (with 72-layer encoders) surpass strong baselines by +2.2~+3.1 BLEU points. In addition, our deep MSC achieves a BLEU score of 30.56 on WMT14 English-German task that significantly outperforms state-of-the-art deep NMT models.

cs.CL↗

arXiv2020

Collaborative Motion Prediction via Neural Motion Message Passing

Motion prediction is essential and challenging for autonomous vehicles and social robots. One challenge of motion prediction is to model the interaction among traffic actors, which could cooperate with each other to avoid collisions or form groups. To address this challenge, we propose neural motion message passing (NMMP) to explicitly model the interaction and learn representations for directed interactions between actors. Based on the proposed NMMP, we design the motion prediction systems for two settings: the pedestrian setting and the joint pedestrian and vehicle setting. Both systems share a common pattern: we use an individual branch to model the behavior of a single actor and an interactive branch to model the interaction between actors, while with different wrappers to handle the varied input formats and characteristics. The experimental results show that both systems outperform the previous state-of-the-art methods on several existing benchmarks. Besides, we provide interpretability for interaction learning.

cs.CV↗

arXiv2020

Magnetic Field Morphology in Interstellar Clouds with the Velocity Gradient Technique

Magnetic fields, while ubiquitous in many astrophysical environments, are challenging to measure observationally. Based on the properties of anisotropy of eddies in magnetized turbulence, the Velocity Gradient Technique is a method synergistic to dust polarimetry that is capable of tracing plane-of-the-sky magnetic field, measuring the magnetization of interstellar media and estimating the fraction of gravitational collapsing gas in molecular clouds using spectral line observations. In this paper, we apply this technique to five low-mass star-forming molecular clouds in the Gould Belt and compare the results to the magnetic-field orientation obtained from polarized dust emission. We find the estimates of magnetic field orientations and magnetization for both methods are statistically similar. We estimate the fraction of collapsing gas in the selected clouds. By means of the Velocity Gradient Technique, we also present the plane-of-the-sky magnetic field orientation and magnetization of the Smith cloud, for which dust polarimetry data are unavailable.

astro-ph.GA↗

arXiv2019

Predictions of Cosmic Microwave Background Foregrounds Dust Polarization Using Velocity Gradients

The observations of fluctuations in the cosmic microwave background provide information about primordial inhomogeneities in the universe. However, the B-mode polarization of the inflationary gravitational wave is contaminated by the Galactic foreground polarized radiation arising from dust aligned by interstellar magnetic fields. To trace magnetic fields we use the Velocity Gradient Technique, which employs a modern understanding of the nature of magneto-hydrodynamics turbulent motions. In this paper, we combine the VGT with the Principal Component Analysis (PCA) to improve the accuracy of magnetic field tracing. We apply the VGT-PCA to the high-resolution neutral hydrogen data from the GALFA-HI survey to predict the polarization of dust. We report that the predicted directions of dust polarization provide good correspondence with those reported by Planck 353GHz, the alignment measure between the two AM=0.79. We show that our results statistically agree with the Planck polarization in terms of magnetic field tracing. We find that the variation of dust emission efficiency across the sky is small. Using our maps of predicted polarization we calculate the ratio of the E and B modes and show that BB/EE=0.53, which is similar to the result from Plank polarization.

astro-ph.CO↗

arXiv2019

A Benchmarking of DCM Based Architectures for Position, Velocity and Torque Controlled Humanoid Robots

This paper contributes towards the benchmarking of control architectures for bipedal robot locomotion. It considers architectures that are based on the Divergent Component of Motion (DCM) and composed of three main layers: trajectory optimization, simplified model control, and whole-body QP control layer. While the first two layers use simplified robot models, the whole-body QP control layer uses a complete robot model to produce either desired positions, velocities, or torques inputs at the joint-level. This paper then compares two implementations of the simplified model control layer, which are tested with position, velocity, and torque control modes for the whole-body QP control layer. In particular, both an instantaneous and a Receding Horizon controller are presented for the simplified model control layer. We show also that one of the proposed architectures allows the humanoid robot iCub to achieve a forward walking velocity of 0.3372 meters per second, which is the highest walking velocity achieved by the iCub robot.

cs.RO↗

arXiv2019

DualVD: An Adaptive Dual Encoding Model for Deep Visual Understanding in Visual Dialogue

Different from Visual Question Answering task that requires to answer only one question about an image, Visual Dialogue involves multiple questions which cover a broad range of visual content that could be related to any objects, relationships or semantics. The key challenge in Visual Dialogue task is thus to learn a more comprehensive and semantic-rich image representation which may have adaptive attentions on the image for variant questions. In this research, we propose a novel model to depict an image from both visual and semantic perspectives. Specifically, the visual view helps capture the appearance-level information, including objects and their relationships, while the semantic view enables the agent to understand high-level visual semantics from the whole image to the local regions. Futhermore, on top of such multi-view image features, we propose a feature selection framework which is able to adaptively capture question-relevant information hierarchically in fine-grained level. The proposed method achieved state-of-the-art results on benchmark Visual Dialogue datasets. More importantly, we can tell which modality (visual or semantic) has more contribution in answering the current question by visualizing the gate values. It gives us insights in understanding of human cognition in Visual Dialogue.

cs.CV↗

arXiv2019

A Unified Phonon Interpretation for the Non-Fourier Heat Conduction by Non-equilibrium Molecular Dynamics Simulations

Nanoconfinement induces many intriguing non-Fourier heat conduction phenomena that have been extensively studied in recent years, such as the nonlinear temperature profile inside the devices, the temperature jumps near the contacts, and the finite-size effects. The understanding of these phenomena, however, has been a matter of debate over the past two decades. In this work, we demonstrate a unified phonon interpretation of non-Fourier heat conduction which can help to understand these phenomena by a mode-to-mode correspondence between the non-equilibrium molecular dynamics (NEMD) simulations and the mode-resolved phonon Boltzmann transport equation (BTE). It is found that the nanoscale phonon transport characteristics including temperature profile, the heat flux value and the modal temperature depend on the applied thermal reservoirs on the two contacts. Our NEMD simulations demonstrate that Langevin thermostat behaves like an infinitely large thermal reservoir and provides thermally equilibrium mode-resolved phonon outlets, while biased reservoirs, e.g., Nose-Hoover chain thermostat and velocity rescaling method behave like non-equilibrium phonon outlets. Our interpretation clearly demonstrates that the non-Fourier heat transport phenomena are originated from a combination of non-diffusive phonon transport and phonon thermal nonequilibrium. This work provides a clear understanding of nanoscale heat transport and may guide the measurement and control of thermal transport in various applications.

cond-mat.mes-hall↗

arXiv2019

Intensity Gradients Technique: Synergy with Velocity Gradients and Polarization Studies

Magnetic fields are ubiquitous in the interstellar medium but are notoriously difficult to study through observation. Making use of the advances in our understanding of MHD turbulence and turbulent reconnection, the Velocity Gradient Technique (VGT) was suggested and successfully applied to study magnetic fields utilizing spectroscopic data. Applying the tools developed for VGT to intensity statistics, we introduce the Intensity Gradients Technique (IGT) as a complementary tool that can be used synergistically with VGT. In this paper, we apply IGT to a diffuse HI region selected from the GALFA-HI survey and compare the intensity gradient maps with those obtained using velocity gradients as well as Planck polarization measurements. We demonstrate the possibility of using IGT and VGT for both studying the magnetic field and identifying shocks in the diffuse interstellar medium. We also explore the ability of IGT in locating self-gravitating regions and calculating Alfvenic Mach number, both alone and in combination with VGT and polarimetry. We compare IGT with the Histogram of Relative Orientation (HRO), which utilizes intensity gradients to characterize the relative orientation of column density structures and local magnetic fields.

astro-ph.GA↗

arXiv2019

Tracing Multi-Scale Magnetic Field Structure Using Multiple Chemical Tracers in Giant Molecular Clouds

Probing magnetic fields in Giant Molecular Clouds is often challenging. Fortunately, recently simulations show that analysis of velocity gradients (the Velocity Gradient Technique) can be used to map out the magnetic field morphology of different physical layers within molecular clouds when applied CO isotopologues with different optical depths. Here, we test the effectiveness of the Velocity Gradient Technique in reconstructing the magnetic field structure of the molecular cloud Vela C, employing seven chemical tracers that have different optical depths, i.e. 12CO, 13CO, C18O, CS, HNC, HCO+, and HCN. Our results show good correspondence between the magnetic field morphology inferred from velocity gradients using these different molecular tracers and the magnetic field morphology inferred from BLASTPol polarization observations. We also explore the possibility of using a combination of velocity gradients for multiple chemical tracers to explain the structure of the magnetic field in molecular clouds. We search for signatures of gravitational collapse in the alignment of the velocity gradients and magnetic field and conclude that collapsing regions constitute a small fraction of the cloud.

astro-ph.GA↗

arXiv2019

Robust Tensor Recovery with Fiber Outliers for Traffic Events

Event detection is gaining increasing attention in smart cities research. Large-scale mobility data serves as an important tool to uncover the dynamics of urban transportation systems, and more often than not the dataset is incomplete. In this article, we develop a method to detect extreme events in large traffic datasets, and to impute missing data during regular conditions. Specifically, we propose a robust tensor recovery problem to recover low rank tensors under fiber-sparse corruptions with partial observations, and use it to identify events, and impute missing data under typical conditions. Our approach is scalable to large urban areas, taking full advantage of the spatio-temporal correlations in traffic patterns. We develop an efficient algorithm to solve the tensor recovery problem based on the alternating direction method of multipliers (ADMM) framework. Compared with existing $l_1$ norm regularized tensor decomposition methods, our algorithm can exactly recover the values of uncorrupted fibers of a low rank tensor and find the positions of corrupted fibers under mild conditions. Numerical experiments illustrate that our algorithm can exactly detect outliers even with missing data rates as high as 40%, conditioned on the outlier corruption rate and the Tucker rank of the low rank tensor. Finally, we apply our method on a real traffic dataset corresponding to downtown Nashville, TN, USA and successfully detect the events like severe car crashes, construction lane closures, and other large events that cause significant traffic disruptions.

eess.SP↗

arXiv2019

Scene Graph Reasoning with Prior Visual Relationship for Visual Question Answering

One of the key issues of Visual Question Answering (VQA) is to reason with semantic clues in the visual content under the guidance of the question, how to model relational semantics still remains as a great challenge. To fully capture visual semantics, we propose to reason over a structured visual representation - scene graph, with embedded objects and inter-object relationships. This shows great benefit over vanilla vector representations and implicit visual relationship learning. Based on existing visual relationship models, we propose a visual relationship encoder that projects visual relationships into a learned deep semantic space constrained by visual context and language priors. Upon the constructed graph, we propose a Scene Graph Convolutional Network (SceneGCN) to jointly reason the object properties and relational semantics for the correct answer. We demonstrate the model's effectiveness and interpretability on the challenging GQA dataset and the classical VQA 2.0 dataset, remarkably achieving state-of-the-art 54.56% accuracy on GQA compared to the existing best model.

cs.MM↗

arXiv2019

GraphNAS: Graph Neural Architecture Search with Reinforcement Learning

Graph Neural Networks (GNNs) have been popularly used for analyzing non-Euclidean data such as social network data and biological data. Despite their success, the design of graph neural networks requires a lot of manual work and domain knowledge. In this paper, we propose a Graph Neural Architecture Search method (GraphNAS for short) that enables automatic search of the best graph neural architecture based on reinforcement learning. Specifically, GraphNAS first uses a recurrent network to generate variable-length strings that describe the architectures of graph neural networks, and then trains the recurrent network with reinforcement learning to maximize the expected accuracy of the generated architectures on a validation data set. Extensive experimental results on node classification tasks in both transductive and inductive learning settings demonstrate that GraphNAS can achieve consistently better performance on the Cora, Citeseer, Pubmed citation network, and protein-protein interaction network. On node classification tasks, GraphNAS can design a novel network architecture that rivals the best human-invented architecture in terms of test set accuracy.

cs.LG↗

arXiv2019

Analyzing Client Behavior in a Syringe Exchange Program

Multiple syringe exchange programs serve the Chicago metropolitan area, providing support for drug users to help prevent infectious diseases. Using data from one program over a ten-year period, we study the behavior of its clients, focusing on the temporal process governing their visits to service locations and on their demographics. We construct a phase-type distribution with an affine relationship between model parameters and features of an individual client. The phase-type distribution governs inter-arrival times between reoccurring visits of each client and is informed by characteristics of a client including age, gender, ethnicity, and more. The inter-arrival time model is a sub-model in a simulation that we construct for the larger system, which allows us to provide a personalized prediction regarding the client's time-to-return to a service location so that better intervention decisions can be made with the help of simulation.

stat.AP↗

arXiv2019

Semantic Modeling of Textual Relationships in Cross-Modal Retrieval

Feature modeling of different modalities is a basic problem in current research of cross-modal information retrieval. Existing models typically project texts and images into one embedding space, in which semantically similar information will have a shorter distance. Semantic modeling of textural relationships is notoriously difficult. In this paper, we propose an approach to model texts using a featured graph by integrating multi-view textual relationships including semantic relations, statistical co-occurrence, and prior relations in the knowledge base. A dual-path neural network is adopted to learn multi-modal representations of information and cross-modal similarity measure jointly. We use a Graph Convolutional Network (GCN) for generating relation-aware text representations, and use a Convolutional Neural Network (CNN) with non-linearities for image representations. The cross-modal similarity measure is learned by distance metric learning. Experimental results show that, by leveraging the rich relational semantics in texts, our model can outperform the state-of-the-art models by 3.4% and 6.3% on accuracy on two benchmark datasets.

cs.MM↗