arXiv subjects

Kun Xu

Publications and source records attributed to Kun Xu.

At least 325 records · Page 18Linked to original sources

arXiv2018

A Unified Gas-kinetic Particle Method for Multiscale Photon Transport

In this work, we present a unified gas-kinetic particle (UGKP) method for the simulation of multiscale photon transport. The multiscale nature of the particle method mainly comes from the recovery of the time evolution flux function in the unified gas-kinetic scheme (UGKS) through a coupled dynamic process of particle transport and collision. This practice improves the original operator splitting approach in the Monte Carlo method, such as the separated treatment of particle transport and collision. As a result, with the variation of the ratio between numerical time step and local photon's collision time, different transport physics can be fully captured in a single computation. In the diffusive limit, the UGKP method could recover the solution of the diffusion equation with the cell size and time step being much larger than the photon's mean free path and the mean collision time. In the free transport limit, it presents an exact particle tracking process as the original Monte Carlo method. In the transition regime, the weights of particle free transport and collision are determined by the ratio of local numerical time step to the photon's collision time. Several one-dimensional numerical examples covering all transport regimes from the optically thin to optically thick are computed to validate the accuracy and efficiency of the current scheme. In comparison with the $S_N$ discrete ordinate method, the UGKP method is based on particles and avoids the discretization of particle velocity space, which does not suffer from the ray effect.

physics.comp-ph↗

arXiv2018

Physical Modeling and Numerical Studies of Three-dimensional Non-equilibrium Multi-temperature Flows

For increasingly rarefied flowfields, the Navier-Stokes (NS) equations lose accuracy partially due to the single temperature approximation. To overcome this barrier, a continuum multi-temperature model based on the Bhatnagar-Gross-Krook (BGK) equation coupled with the Landau-Teller-Jeans relaxation model has been proposed for two-dimensional hypersonic non-equilibrium multi-temperature flow computation. In recent study, a two-stage fourth-order gas-kinetic scheme (GKS) has been developed for equilibrium flows, which achieves a fourth-order accuracy in space and time as well as high efficiency and robustness. In this paper, targeting for accurate and efficient simulation of multi-temperature non-equilibrium flows, a high-order three-dimensional multi-temperature GKS is implemented under the two-stage fourth-order framework, with the fourth-order Simpson interpolation rule for the newly emerged source term. Simulations on decaying homogeneous isotropic turbulence, low-density nozzle flow, rarefied hypersonic flow over a flat plate, and type IV shock-shock interaction are used to validate the multi-temperature model through the comparison with experimental measurements. The unified gas kinetic scheme (UGKS) results, and the Direct simulation Monte Carlo (DSMC) solutions will be used as well in some cases for validation. Computational results not only confirm the high-order accuracy and quite robustness of this scheme, but also show the significant improvement on computational efficiency compared with UGKS and DSMC, especially in the near continuum flow regime.

physics.flu-dyn↗

arXiv2018

QCD critical end point from a realistic PNJL model

With parameters fixed by critical temperature and equation of state at zero baryon chemical potential, a realistic Polyakov--Nambu--Jona-Lasinio (rPNJL) model predicts a critical end point of chiral phase transition at $(μ_B^E= 720 {\rm MeV}, T^E=93 {\rm MeV})$. The extracted freeze-out line from heavy ion collisions is close to the chiral phase transition boundary in the rPNJL model, and the kurtosis $κσ^2$ of baryon number fluctuations from the rPNJL model along the experimental freeze-out line agrees well with the BES-I measurement. Our analysis shows that the dip structure of measured $κσ^2$ is determined by the relationship between the freeze-out line and chiral phase transition line at low baryon density region, and the peak structure can be regarded as a clean signature for the existence of CEP.

hep-ph↗

arXiv2018

A Unified Gas-kinetic Scheme for Continuum and Rarefied Flows VI: Dilute Disperse Gas-Particle Multiphase System

In this paper, a unified gas kinetic scheme for multiphase dilute gas-particle system is proposed. The UGKS multiphase (UGKS-M) is a finite volume method, which captures flow physics in the regimes from collisionless multispecies transport to the two-fluid hydrodynamic Navier-Stokes (NS) solution with the variation of Knudsen number, and from granular flow regime to dusty gas dynamics with the variation of Stokes number. The main reason for preserving the multiscale nature in UGKS-M is mainly coming from the direct modeling of the flow physics in the scales of discrete cell size and time step, where the ratio of the time step over the particle collision time determines flow behavior in different regimes. For the particle phase, the integral solution of the kinetic equation is used in the construction of the numerical flux, which takes into account the particle transport, collision, and acceleration. The gas phase, which is assumed to be in the continuum flow regime, evolves numerically by the gas kinetic scheme (GKS), which is a subset of the UGKS for the Navier-Stokes (NS) solutions. The interaction between the gas and particle phase is calculated based on a velocity space mapping method, which solves accurately the kinetic acceleration process. The stability of UGKS-M is determined by the CFL condition only. With the inclusion of the material temperature evolution equation of solid particles, the UGKS-M conserves the total mass, momentum, and energy for the whole multiphase system. In the numerical tests, the UGKS-M shows good multiscale property in capturing the particle trajectory crossing (PTC), particle wall reflecting phenomena, and vortex-induced segregation of inertial particles under different Stokes numbers. The scheme is also applied to simulate shock induced fluidization problem and the simulation results agree well with experimental measurement.

physics.comp-ph↗

arXiv2018

Exploiting Rich Syntactic Information for Semantic Parsing with Graph-to-Sequence Model

Existing neural semantic parsers mainly utilize a sequence encoder, i.e., a sequential LSTM, to extract word order features while neglecting other valuable syntactic information such as dependency graph or constituent trees. In this paper, we first propose to use the \textit{syntactic graph} to represent three types of syntactic information, i.e., word order, dependency and constituency features. We further employ a graph-to-sequence model to encode the syntactic graph and decode a logical form. Experimental results on benchmark datasets show that our model is comparable to the state-of-the-art on Jobs640, ATIS and Geo880. Experimental results on adversarial examples demonstrate the robustness of the model is also improved by encoding more syntactic information.

cs.CL↗

arXiv2018

Deep Structured Generative Models

Deep generative models have shown promising results in generating realistic images, but it is still non-trivial to generate images with complicated structures. The main reason is that most of the current generative models fail to explore the structures in the images including spatial layout and semantic relations between objects. To address this issue, we propose a novel deep structured generative model which boosts generative adversarial networks (GANs) with the aid of structure information. In particular, the layout or structure of the scene is encoded by a stochastic and-or graph (sAOG), in which the terminal nodes represent single objects and edges represent relations between objects. With the sAOG appropriately harnessed, our model can successfully capture the intrinsic structure in the scenes and generate images of complicated scenes accordingly. Furthermore, a detection network is introduced to infer scene structures from a image. Experimental results demonstrate the effectiveness of our proposed method on both modeling the intrinsic structures, and generating realistic images.

cs.LG↗

arXiv2018

Cell Size Effect on Computational Fluid Dynamics: The Limitation Principle for Flow Simulation

For theoretical gas dynamics, the flow regimes are classified according to the Knudsen number. For computational fluid dynamics (CFD), the numerical flow field is the projection of the physical flow field onto the discrete space and time, which is related to the cell Knudsen number. The real representable flow regimes are controlled by these two parameters. According to the values of Knudsen number and cell Knudsen number, we study the classification of the numerical flow regimes. In the process of mesh refinement, the numerical experiments show the change of numerical flow regime from continuum, to near-continuum, and to non-equilibrium one. The change of flow regime with different cell resolution is the limitation principle for the numerical simulation, which is the best a multiscale method can do. In other words, we should have changeable numerical governing equations in different mesh size scale, and they are coupled with the traditional physical equations in different scales, such as the Navier-Stokes and Boltzmann. Under the multiscale modeling, a mesh refinement is a process in resolving the flow physics in different scale. The verification and validation (V$\&$V) need include the physical modeling mechanism in the mesh refinement process. The traditional idea of mesh refinement for targeting a fixed partial differential equation cannot achieve the final goal of computation, which is to recover the flow physics as truthfully as possible under the limitation of the cell resolution.

physics.comp-ph↗

arXiv2018

Chinese Text in the Wild

We introduce Chinese Text in the Wild, a very large dataset of Chinese text in street view images. While optical character recognition (OCR) in document images is well studied and many commercial tools are available, detection and recognition of text in natural images is still a challenging problem, especially for more complicated character sets such as Chinese text. Lack of training data has always been a problem, especially for deep learning methods which require massive training data. In this paper we provide details of a newly created dataset of Chinese text with about 1 million Chinese characters annotated by experts in over 30 thousand street view images. This is a challenging dataset with good diversity. It contains planar text, raised text, text in cities, text in rural areas, text under poor illumination, distant text, partially occluded text, etc. For each character in the dataset, the annotation includes its underlying character, its bounding box, and 6 attributes. The attributes indicate whether it has complex background, whether it is raised, whether it is handwritten or printed, etc. The large size and diversity of this dataset make it suitable for training robust neural networks for various tasks, particularly detection and recognition. We give baseline results using several state-of-the-art networks, including AlexNet, OverFeat, Google Inception and ResNet for character recognition, and YOLOv2 for character detection in images. Overall Google Inception has the best performance on recognition with 80.5% top-1 accuracy, while YOLOv2 achieves an mAP of 71.0% on detection. Dataset, source code and trained models will all be publicly available on the website.

cs.CV↗

arXiv2018

A velocity-space adaptive unified gas kinetic scheme for continuum and rarefied flows

In this paper, a unified gas kinetic scheme with adaptive velocity space (AUGKS) for multiscale flow transport will be developed. In near-equilibrium flow regions, particle distribution function is close to the Chapman-Enskog expansion and can be formulated with a continuous velocity space, where only macroscopic conservative variables are updated. With the emerging of non-equilibrium effects, the AUGKS automatically switches to a discrete velocity space to follow the evolution of particle distribution function. Based on the Chapman-Enskog expansion, a criterion is proposed in this paper to quantify the intensity of non-equilibrium effects and is used for the continuous-discrete velocity space transformation. Following the scale-dependent local evolution solution, the AUGKS presents the discretized gas dynamic equations directly on the cell size and time step scales, i.e., the so-called direct modeling method. As a result, the scheme is able to capture the cross-scale flow physics from particle transport to hydrodynamic wave propagation, and provides a continuous variation of solutions from the Boltzmann to the Navier-Stokes. Under the unified framework, different from conventional DSMC-NS hybrid method, the AUGKS does not need a buffer zone to match up kinetic and hydrodynamic solutions. Instead, a continuous and discrete particle velocity space is naturally connected, which is feasible for the numerical simulations with unsteadiness or complex geometries. Compared with the asymptotic preserving (AP) methods which solves kinetic equations uniformly over the entire flow field with discretized velocity space, the current velocity-space adaptive unified scheme speeds up the computation and reduces the memory requirement in multiscale flow problems, and maintains the equivalent accuracy. The AUGKS provides an effective tool for non-equilibrium flow simulations.

physics.comp-ph↗

arXiv2018

Two-Stage Fourth-order Gas-kinetic Scheme for Three-dimensional Euler and Navier-Stokes Solutions

For the one-stage third-order gas-kinetic scheme (GKS), success applications have been achieved for the three-dimensional compressible flow computations [33]. The high-order accuracy of the scheme is obtained directly by integrating a multidimensional time-accurate gas distribution function over the cell interface within a time step without implementing Gaussian quadrature points and Runge-Kutta time-stepping technique. However, for the further increasing the order of the scheme, such as the fourth-order one, the formulation becomes very complicated for the multidimensional flow. Recently, a two-stage fourth-order GKS with high efficiency has been constructed for two-dimensional inviscid and viscous flow computations [22,32], and the scheme uses the time accurate flux function and its time derivatives. In this paper, a fourth-order GKS is developed for the threedimensional flows under the two-stage framework. Based on the three-dimensional WENO reconstruction and flux evaluation at Gaussian quadrature points on a cell interface, the high-order accuracy in space is achieved first. Then, the two-stage time stepping method provides the high accuracy in time. In comparison with the formal third-order GKS [33], the current fourth-order method not only improves the accuracy of the scheme, but also reduces the complexity of the gas-kinetic solver greatly. More importantly, the fourth-order GKS has the same robustness as the second-order shock capturing scheme. This scheme is applied to both inviscid and viscous, and low and high speed flow computations. Numerical results validate the outstanding reliability and applicability of the scheme for three-dimensional flows, such as turbulent one.

math.NA↗

arXiv2018

An Implicit Unified Gas-kinetic Scheme for Unsteady Flow in All Knudsen Regimes

The unified gas-kinetic scheme (UGKS) is a direct modeling method for multiple scale transports. The modeling scale of the scheme is the mesh size and time step, and the ratios of the mesh size over particle mean free path or the time step over particle collision time determine the local evolution regime of flow physics. For unsteady flow computation, due to the CFL condition the time step in the explicit UGKS is limited by the smallest cell size in the computational domain. As a result, for a largely stretched non-uniform mesh the global time step becomes very small and the ratio of the time step over local particle collision time may get a very small value. Under such a circumstance, the physics in explicit UGKS may be constrained by the kinetic scale physics only. The multiscale UGKS reduces into a single scale discrete velocity method where the free transport is used for the flux evaluation. In order to keep the advantages of UGKS and improve the efficiency for unsteady flow computation, the restriction from global CFL condition on the time step has to be removed. In this paper, we develop an implicit UGKS (IUGKS) for unsteady flow by alternatively solving the macroscopic and microscopic governing equations within a step iteratively. In order to preserve the coherent flow evolution and multiscale nature, the numerical flux is still evaluated by the explicit UGKS with the local CFL condition. Therefore, the multiscale property of the UGKS modeling is preserved over non-uniform meshes. Many numerical examples are included to validate the scheme for both continuum and rarefied flows. The IUGKS presents reasonably good results with second order accuracy for unsteady flow computation and the efficiency has been improved by dozen of times in comparison with the explicit UGKS.

physics.comp-ph↗

arXiv2017

A Compact Fourth-order Gas-kinetic Scheme for the Euler and Navier-Stokes Solutions

In this paper, a fourth-order compact gas-kinetic scheme (GKS) is developed for the compressible Euler and Navier-Stokes equations under the framework of two-stage fourth-order temporal discretization and Hermite WENO (HWENO) reconstruction. Due to the high-order gas evolution model, the GKS provides a time dependent gas distribution function at a cell interface. This time evolution solution can be used not only for the flux evaluation across a cell interface and its time derivative, but also time accurate evolution solution at a cell interface. As a result, besides updating the conservative flow variables inside each control volume, the GKS can get the cell averaged slopes inside each control volume as well through the differences of flow variables at the cell interfaces. So, with the updated flow variables and their slopes inside each cell, the HWENO reconstruction can be naturally implemented for the compact high-order reconstruction at the beginning of next step. Therefore, a compact higher-order GKS, such as the two-stages fourth-order compact scheme can be constructed. This scheme is as robust as second-order one, but more accurate solution can be obtained. In comparison with compact fourth-order DG method, the current scheme has only two stages instead of four within each time step for the fourth-order temporal accuracy, and the CFL number used here can be on the order of $0.5$ instead of $0.11$ for the DG method. Through this research, it concludes that the use of high-order time evolution model rather than the first order Riemann solution is extremely important for the design of robust, accurate, and efficient higher-order schemes for the compressible flows.

physics.comp-ph↗

arXiv2017

Triple Generative Adversarial Nets

Generative Adversarial Nets (GANs) have shown promise in image generation and semi-supervised learning (SSL). However, existing GANs in SSL have two problems: (1) the generator and the discriminator (i.e. the classifier) may not be optimal at the same time; and (2) the generator cannot control the semantics of the generated samples. The problems essentially arise from the two-player formulation, where a single discriminator shares incompatible roles of identifying fake samples and predicting labels and it only estimates the data without considering the labels. To address the problems, we present triple generative adversarial net (Triple-GAN), which consists of three players---a generator, a discriminator and a classifier. The generator and the classifier characterize the conditional distributions between images and labels, and the discriminator solely focuses on identifying fake image-label pairs. We design compatible utilities to ensure that the distributions characterized by the classifier and the generator both converge to the data distribution. Our results on various datasets demonstrate that Triple-GAN as a unified model can simultaneously (1) achieve the state-of-the-art classification results among deep generative models, and (2) disentangle the classes and styles of the input and transfer smoothly in the data space via interpolation in the latent space class-conditionally.

cs.LG↗

arXiv2017

An investigation of non-equilibrium heat transport in a gas system under external force field

The gas dynamics under external force field is essentially associated with multiple scale nature due to the large variations of density and local Knudsen number. Single scale fluid dynamic equations, such as the Boltzmann and Navier-Stokes equations, are valid in their respective modeling scales, and it is challenging for the modeling and computation of a multiple scale problem across different regimes and capture the corresponding non-equilibrium flow physics. Based on the direct modeling of conservation laws in the discretized space, a well-balanced unified gas-kinetic scheme (UGKS) for multiscale flow transport under external force field has been developed and is used in the current study of non-equilibrium gaseous flow under external force field. With the variation of modeling scale, i.e., the cell size and time step, the UGKS is able to recover cross-scale flow physics from particle transport to hydrodynamic wave propagation. Theoretical analysis based on the kinetic model equation is presented to conceptually illustrate the effects of external force on the non-equilibrium heat transport.

physics.flu-dyn↗

arXiv2017

High-resolution and Fully-programmable Microwave-shaper

We here propose and demonstrate a point-by-point programmable broadband microwave spectrum processor with high-resolution up to tens of MHz. We achieve this by bandwidth-minified mapping a programmable optical spectrum processor, which has much larger bandwidth and lower frequency resolution, to a microwave one, and make sure the mapping is a similarity transformation. As a comparison, the traditional optical-to-microwave mapping based on super-heterodyne is bandwidth-preserved and identical. Here the optical spectrum processor is firstly sliced by a high-quality-factor optical resonator, which has periodic transmission peak with MHz-level bandwidth, and then is mapped to microwave domain by optical-frequency-comb-assisted multi-heterodyne. We demonstrate the high frequency resolution and full function reconfigurability experimentally. In a numerical example, we show that the group delay variation of optical spectrum processor could be greatly enlarged after mapping. The resolution improvement and group delay magnification distinguish our proposal significantly from previous spectrum mapping from optics to microwave.

physics.optics↗

arXiv2017

A Family of High-order Gas-kinetic Schemes and Its Comparison with Riemann Solver Based High-order Methods

Most high order computational fluid dynamics (CFD) methods for compressible flows are based on Riemann solver for the flux evaluation and Runge-Kutta (RK) time stepping technique for temporal accuracy. The main advantage of this kind of approach is the easy implementation and stability enhancement by introducing more middle stages. However, the nth-order time accuracy needs no less than n stages for the RK method, which is very time and memory consuming for a high order method. On the other hand, the multi-stage multi-derivative (MSMD) method can be used to achieve the same order of time accuracy using less middle stages, once the time derivatives of the flux function is used.The gas kinetic scheme (GKS) provides such a time accurate evolution model. By combining the second-order or third-order GKS flux functions with the MSMD technique, a family of high order gas kinetic methods can be constructed. As an extension of the previous 2-stage 4th-order GKS, the 5th-order schemes with 2 and 3 stages will be developed in this paper. Based on the same 5th-order WENO reconstruction, the performance of gas kinetic schemes from the 2nd- to the 5th-order time accurate methods will be evaluated. The results show that the 5th-order scheme can achieve the theoretical order of accuracy for the Euler equations, and present accurate Navier-Stokes solutions as well due to the coupling of inviscid and viscous terms in the GKS formulation. In comparison with Riemann solver based 5th-order RK method, the high order GKS has advantages in terms of efficiency, accuracy, and robustness, for all test cases. It provides a promising direction for the development of high-order CFD methods for the computation of complex flows, such as turbulence and acoustics with shock interactions.

physics.comp-ph↗

arXiv2017

The YouTube-8M Kaggle Competition: Challenges and Methods

We took part in the YouTube-8M Video Understanding Challenge hosted on Kaggle, and achieved the 10th place within less than one month's time. In this paper, we present an extensive analysis and solution to the underlying machine-learning problem based on frame-level data, where major challenges are identified and corresponding preliminary methods are proposed. It's noteworthy that, with merely the proposed strategies and uniformly-averaging multi-crop ensemble was it sufficient for us to reach our ranking. We also report the methods we believe to be promising but didn't have enough time to train to convergence. We hope this paper could serve, to some extent, as a review and guideline of the YouTube-8M multi-label video classification benchmark, inspiring future attempts and research.

cs.CV↗

arXiv2017

Frequency-oriented sub-sampling by photonic Fourier transform and I/Q demodulation

Sub-sampling can acquire directly a passband within a broad radio frequency (RF) range, avoiding down-conversion and low-phase-noise tunable local oscillation (LO). However, sub-sampling suffers from band folding and self-image interference. In this paper we propose a frequency-oriented sub-sampling to solve the two problems. With ultrashort optical pulse and a pair of chromatic dispersions, the broadband RF signal is firstly short-time Fourier-transformed to a spectrum-spread pulse. Then a time slot, corresponding to the target spectrum slice, is coherently optical-sampled with in-phase/quadrature (I/Q) demodulation. We demonstrate the novel bandpass sampling by a numerical example, which shows the desired uneven intensity response, i.e. pre-filtering. We show in theory that appropriate time-stretch capacity from dispersion can result in pre-filtering bandwidth less than sampling rate. Image rejection due to I/Q sampling is also analyzed. A proof-of-concept experiment, which is based on a time-lens sampling source and chirped fiber Bragg gratings (CFBGs), shows the center-frequency-tunable pre-filtered sub-sampling with bandwidth of 6 GHz around, as well as imaging rejection larger than 26 dB. Our technique may benefit future broadband RF receivers for frequency-agile Radar or channelization.

physics.ins-det↗