arXiv subjects

Kun Xu

Publications and source records attributed to Kun Xu.

At least 145 records · Page 8Linked to original sources

arXiv2024

Robust characterization of photonic integrated circuits

Photonic integrated circuits (PICs) offer ultra-broad optical bandwidths that enable unprecedented data throughputs for signal processing applications. Dynamic reconfigurability enables compensation of fabrication flaws and fluctuating external environments, tuning for adaptive equalization and training of optical neural networks. The initial step in PIC reconfiguration entails measuring its dynamic performance, often described by its frequency response. While measuring the amplitude response is straightforward, e.g. using a tunable laser and optical power meter, measuring the phase response presents challenges due to various factors, including phase variations in test connections and instrumentation limitations. To address these challenges, a universal and robust characterization technique is proposed, which uses an on-chip reference path coupled to the signal processing core (SPC), with a delay larger or smaller than the total delay across the signal processing paths. A Fourier transform of the chip's power response reveals the SPC's impulse response. The method is more robust against low reference-path power and imprecise delays. Experiments using a finite-impulse-response (FIR) structure demonstrate rapid SPC training, overcoming thermal crosstalk and device imperfections. This approach offers a promising solution for PIC characterization, facilitating expedited physical parameter training for advanced applications in communications and optical neural networks.

physics.optics↗

arXiv2024

An implicit adaptive unified gas-kinetic scheme for steady-state solutions of non-equilibrium flows

Nonequilibrium flows have been frequently encountered in various aerospace engineering applications. To understand nonequilibrium physics, multiscale effects, and the dynamics in these applications, an effective and reliable multiscale scheme for all flow regimes is required. Following the direct modeling methodology, the adaptive unified gas-kinetic scheme employs discrete velocity space (DVS) to accurately capture the non-equilibrium physics, recovering the original unified gas-kinetic scheme (UGKS), and adaptively employs continuous distribution functions based on the Chapman-Enskog expansion to achieve better efficiency. Different regions are dynamically coupled at the cell interface through the fluxes from the discrete and continuous gas distribution functions, thereby avoiding any buffer zone between them. In the current study, an implicit adaptive unified gas-kinetic scheme (IAUGKS) is constructed to further enhance the efficiency of steady-state solutions. The current scheme employs implicit macroscopic governing equations and couples them with implicit microscopic governing equations within the non-equilibrium region, resulting in high convergence efficiency in all flow regimes. A series of numerical tests were conducted for high Mach number flows around diverse geometries such as a cylinder, a sphere, an X-38-like vehicle, and a space station. The current scheme can capture the non-equilibrium physics and provide accurate predictions of surface quantities. In comparison with the original UGKS, the velocity space adaptation, unstructured DVS, and implicit iteration significantly improve the efficiency by one or two orders of magnitude. Given its exceptional efficiency and accuracy, the IAUGKS serves as an effective tool for nonequilibrium flow simulations.

physics.flu-dyn↗

arXiv2024

A Memory Reduction Compact Gas Kinetic Scheme on 3D Unstructured Meshes

This paper introduces a memory-reduction third-order compact gas-kinetic scheme (CGKS) for solving compressible Euler and Navier-Stokes equations on 3D unstructured meshes. The scheme utilizes a time-evolution gas distribution function to provide a time-evolution solution at cell interfaces, enabling the implementation of Hermite WENO techniques for high-order reconstruction. However, the HWENO method needs to store a coefficients matrix for the quadratic polynomial to achieve third-order accuracy, resulting in high memory usage. A novel reconstruction method, built upon HWENO reconstruction, has been designed to enhance computational efficiency and reduce memory usage compared to the original CGKS. The simple idea is that the first-order and second-order terms of the quadratic polynomials are determined in a two-step way. In the first step, the second-order terms are obtained from the reconstruction of a linear polynomial of the first-order derivatives by only using the cell-averaged slopes, since the second-order derivatives are nothing but the "derivatives of derivatives". Subsequently, the first-order terms left can be determined by the linear reconstruction only using cell-averaged values. Thus, we successfully split one quadratic least-square regression into several linear least-square regressions, which are commonly used in a second-order finite volume code. Since only a small matrix inversion is needed in a 3-D linear least-square regression, the computational cost for the new reconstruction is dramatically reduced and the storage of the reconstruction-coefficient matrix is no longer necessary. The proposed new reconstruction technique can reduce the overall computational cost by about 20 to 30 percent. The challenging large-scale unsteady numerical simulation is performed, which demonstrates that the current improvement brings the CGKS to a new level for industrial applications.

math.NA↗

arXiv2024

PAC.V. The Roles of Mass and Environment in the Quenching of Galaxies

The roles that mass and environment play in the galaxy quenching are still under debate. Leveraging the Photometric objects Around Cosmic webs (PAC) method, we analyze the excess surface distribution $\bar{n}_2w_{\rm{p}}(r_{\rm{p}})$ of photometric galaxies in different color (rest-frame $u-r$) within the stellar mass range of $10^{9.0}M_{\odot}\sim10^{11.0}M_{\odot}$ around spectroscopic massive central galaxies ($10^{10.9}\sim10^{11.7}M_{\odot}$) at the redshift interval $0<z_s<0.7$, utilizing data from the Hyper SuprimeCam Subaru Strategic Program and the spectroscopic samples of Slogan Digital Sky Survey (i.e. Main, LOWZ and CMASS samples). We find that both mass and environment quenching contribute to the evolution of companion galaxies. To isolate the environment effect, we quantify the quenched fraction excess (QFE) of companion galaxies encircling massive central galaxies within $0.01h^{-1}{\rm{Mpc}}<r_{\rm{p}}<20h^{-1}\rm{Mpc}$, representing the surplus quenched fraction relative to the average. We find that the high density halo environment affects the star formation quenching up to about three times of the virial radius, and this effect becomes stronger at lower redshift. We also find that even after being scaled by the virial radius, the environment quenching efficiency is higher for more massive halos or for companion galaxies of higher stellar mass, though the trends are quite weak. We present a fitting formula that comprehensively captures the QFE across central and companion stellar mass bins, halo-centric distance bins, and redshift bins, offering a valuable tool for constraining galaxy formation models. Furthermore, we have made a quantitative comparison with Illustris-TNG that underscores some important differences, particularly in the excessive quenching of low-mass companion galaxies ($<10^{9.5}M_{\odot}$) by TNG.

astro-ph.GA↗

arXiv2024

Unified Gas-Kinetic Wave-Particle Method for Multiscale Flow Simulation of Partially Ionized Plasma

The Unified Gas-Kinetic Wave-Particle (UGKWP) method is constructed for partially ionized plasma (PIP). This method possesses both multiscale and unified preserving (UP) properties. The multiscale property allows the method to capture a wide range of plasma physics, from the particle transport in the kinetic regime to the two-fluid and magnetohydrodynamics (MHD) in the near continuum regimes, with the variation of local cell Knudsen number and normalized Larmor radius.The unified preserving property ensures that the numerical time step is not limited by the particle collision time in the continuum regime for the capturing of dissipative macroscopic solutions of the resistivity, Hall-effect, and all the way to the ideal MHD equations.The UGKWP is clearly distinguishable from the classical single scale Particle-in-Cell/Monte Carlo Collision (PIC/MCC) methods.The UGKWP method combines the evolution of microscopic velocity distribution with the evolution of macroscopic mean field quantities, granting it UP properties. Moreover, the time step in UGKWP is not constrained by the plasma cyclotron period through the Crank-Nicolson scheme for fluid and electromagnetic field interactions. The momentum and energy exchange between different species is approximated by the Andries-Aoki-Perthame (AAP) model. Overall, the UGKWP method enables a smooth transition from the PIC method in the rarefied regime to the MHD solvers in the continuum regime. This method has been extensively tested on a variety of phenomena ranging from kinetic Landau damping to the macroscopic flow problems, such as the Brio-Wu shock tube, Orszag-Tang vortex, and Geospace Environmental Modeling (GEM) magnetic reconnection. These tests demonstrate that the proposed method can capture the fundamental features of PIP across different scales seamlessly.

physics.comp-ph↗

arXiv2024

Particle systems for mean reflected BSDEs with jumps

In this paper, we study the mean reflected backward stochastic differential equations with jump (BSDEJs). We extend the work of Briand and Hibon on the propagation of chaos for mean reflected BSDEs \cite{briand2021particles} to the jump framework. Besides, we study the reflections for the particle system and obtain the rate of of convergence of the particle system towards the deterministic flat solution to the mean reflected BSDEJ.

math.PR↗

arXiv2024

On the Interacting/Active Lifetime of Supernova Fallback Disk around Isolated Neutron Stars

The fallback disk model is widely accepted to explain long-period neutron stars (NSs) which can't be simulated by magnetic dipole radiation. However, no confirmed detection of disk was found from the newly discovered long period pulsars GLEAM-X 162759.5-523504.3, GPM J1839-10 and the known slowest isolated NSs 1E 161348-5055. This might be that the disks have either been in noninteracting/inactive state where its emission is too weak to be detected or have been disrupted. In this work, we conduct simulations to examine the lifetime of supernova fallback disks around isolated neutron stars. We assume that the disk's mass varies in a self-similar way and its interaction with the NS occurs only in interacting/active state. Our results reveal that nearly all the interacting lifetimes for the disk are shorter than 0.1 Myr while the existence lifetimes are considerably longer.

astro-ph.HE↗

arXiv2024

Video-LaVIT: Unified Video-Language Pre-training with Decoupled Visual-Motional Tokenization

In light of recent advances in multimodal Large Language Models (LLMs), there is increasing attention to scaling them from image-text data to more informative real-world videos. Compared to static images, video poses unique challenges for effective large-scale pre-training due to the modeling of its spatiotemporal dynamics. In this paper, we address such limitations in video-language pre-training with an efficient video decomposition that represents each video as keyframes and temporal motions. These are then adapted to an LLM using well-designed tokenizers that discretize visual and temporal information as a few tokens, thus enabling unified generative pre-training of videos, images, and text. At inference, the generated tokens from the LLM are carefully recovered to the original continuous pixel space to create various video content. Our proposed framework is both capable of comprehending and generating image and video content, as demonstrated by its competitive performance across 13 multimodal benchmarks in image and video understanding and generation. Our code and models are available at https://video-lavit.github.io.

cs.CV↗

arXiv2024

Semantic Importance-Aware Communications with Semantic Correction Using Large Language Models

Semantic communications, a promising approach for agent-human and agent-agent interactions, typically operate at a feature level, lacking true semantic understanding. This paper explores understanding-level semantic communications (ULSC), transforming visual data into human-intelligible semantic content. We employ an image caption neural network (ICNN) to derive semantic representations from visual data, expressed as natural language descriptions. These are further refined using a pre-trained large language model (LLM) for importance quantification and semantic error correction. The subsequent semantic importance-aware communications (SIAC) aim to minimize semantic loss while respecting transmission delay constraints, exemplified through adaptive modulation and coding strategies. At the receiving end, LLM-based semantic error correction is utilized. If visual data recreation is desired, a pre-trained generative artificial intelligence (AI) model can regenerate it using the corrected descriptions. We assess semantic similarities between transmitted and recovered content, demonstrating ULSC's superior ability to convey semantic understanding compared to feature-level semantic communications (FLSC). ULSC's conversion of visual data to natural language facilitates various cognitive tasks, leveraging human knowledge bases. Additionally, this method enhances privacy, as neither original data nor features are directly transmitted.

eess.SP↗

arXiv2024

Photometric Objects Around Cosmic Webs (PAC). VI. High Satellite Fraction of Quasars

The Photometric objects Around Cosmic webs (PAC) approach developed in Xu et al. (2022b) has the advantage of making full use of spectroscopic and deeper photometric surveys. With the merits of PAC, the excess surface density $\bar{n}_2w_{\rm{p}}$ of neighboring galaxies can be measured down to stellar mass $10^{10.80}\,M_{\odot}$ around quasars at redshift $0.8<z_{\rm{s}}<1.0$, with the data from the Sloan Digital Sky Survey IV (SDSS-IV) extended Baryon Oscillation Spectroscopic Survey (eBOSS) and the Dark Energy Spectroscopic Instrument (DESI) Legacy Imaging Surveys. We find that $\bar{n}_2w_{\rm{p}}$ generally increases quite steeply with the decrease of the separation. Using subhalo abundance matching method, we can accurately model the $\bar{n}_2w_{\rm{p}}$ both on small and large scales. We show that the steep increase of the $\bar{n}_2w_{\rm{p}}$ towards the quasars requires that a large fraction $f_{\mathrm{sate}}=0.29_{-0.06}^{+0.05}$ of quasars should be satellites in massive halos, and find that this fraction measurement is insensitive to the assumptions of our modeling. This high satellite fraction indicates that the subhalos have nearly the same probability to host quasars as the halos for the same (infall) halo mass, and the large scale environment has negligible effect on the quasar activity. We show that even with this high satellite fraction, each massive halo on average does not host more than one satellite quasar due to the sparsity of quasars.

astro-ph.GA↗

arXiv2024

Comparison of the high-order Runge-Kutta discontinuous Galerkin method and gas-kinetic scheme for inviscid compressible flow simulations

The Runge--Kutta discontinuous Galerkin (RKDG) method is a high-order technique for addressing hyperbolic conservation laws, which has been refined over recent decades and is effective in handling shock discontinuities. Despite its advancements, the RKDG method faces challenges, such as stringent constraints on the explicit time-step size and reduced robustness when dealing with strong discontinuities. On the other hand, the Gas-Kinetic Scheme (GKS) based on a high-order gas evolution model also delivers significant accuracy and stability in solving hyperbolic conservation laws through refined spatial and temporal discretizations. Unlike RKDG, GKS allows for more flexible CFL number constraints and features an advanced flow evolution mechanism at cell interfaces. Additionally, GKS' compact spatial reconstruction enhances the accuracy of the method and its ability to capture stable strong discontinuities effectively. In this study, we conduct a thorough examination of the RKDG method using various numerical fluxes and the GKS method employing both compact and non-compact spatial reconstructions. Both methods are applied under the framework of explicit time discretization and are tested solely in inviscid scenarios. We will present numerous numerical tests and provide a comparative analysis of the outcomes derived from these two computational approaches.

math.NA↗

arXiv2024

Exponential growth BSDE driven by a marked point process

In this study, we investigate the well-posedness of exponential growth backward stochastic differential equations (BSDEs) driven by a marked point process (MPP) under unbounded terminal conditions. Our analysis utilizes a fixed-point argument, the $θ$-method, and an approximation procedure. Additionally, we establish the solvability of mean-reflected exponential growth BSDEs driven by the MPP using the $θ$-method.

math.PR↗

arXiv2024

Evolutionary Origin of Ultra-long Period Radio Transients

Recently, it discovered two ultra-long period radio transients GLEAM-X J162759.5-523504.3 (J1627) and GPM J1839$-$10 (J1839) with spin periods longer than 1000 s. The origin of these two ultra-long period radio transients is intriguing in understanding the spin evolution of neutron stars (NSs). In this work, we diagnose whether the interaction between strong magnetized NSs and fallback disks can spin NSs down to the observed ultra-long period. Our simulations found that the magnetar+fallback disk model can account for the observed period, period derivative, and X-ray luminosity of J1627 in the quasi-spin-equilibrium stage. To evolve to the current state of J1627, the initial mass-accretion rate of the fallback disk and the magnetic field of the NS are in the range of $(1.1-30)\times10^{24}~\rm g\,s^{-1}$ and $(2-5)\times10^{14}~\rm G$, respectively. In an active lifetime of fallback disk, J1839 is impossible to achieve the observed upper limit of period derivative. Therefore, we propose that J1839 may be in the second ejector phase after the fallback disk becomes inactive. Those NSs with a magnetic field of $(2-6)\times10^{14}~\rm G$ and a fallback disk with an initial mass-accretion rate of $\sim10^{24}-10^{26}~\rm g\,s^{-1}$ are the possible progenitors of J1839.

astro-ph.HE↗

arXiv2024

Net proton number cumulants from viscous hydro with equation of state including a critical end point

In the SMASH-CLVisc-hybrid framework, including SMASH for the initial conditions and the hadronic rescattering stage, and CLVisc for the quark gluon plasma (QGP) evolution, we investigate net baryon number fluctuations via considering the equation of state (EoS) with and without a critical end point (CEP) in the QCD phase transition. Specifically, two distinct QCD EoS are utilized: one with smooth crossover derived from NEOS and another with a critical end point sourced from the rPNJL model. Our results show that non-monotonic behavior of $κσ^2$ is not observed above the collision energy $7.7 {\rm GeV}$, nor are there explicit differences between the EoS characterized by crossover and that by CEP. This could be attributed to the significant deviation of the freeze-out line from the location of CEP. It is also found that the pure SMASH result of $κσ^2$ is positive and close to zero at 3 GeV, which is different from the negative value observed from STAR.

hep-ph↗

arXiv2024

Unified Language-Vision Pretraining in LLM with Dynamic Discrete Visual Tokenization

Recently, the remarkable advance of the Large Language Model (LLM) has inspired researchers to transfer its extraordinary reasoning capability to both vision and language data. However, the prevailing approaches primarily regard the visual input as a prompt and focus exclusively on optimizing the text generation process conditioned upon vision content by a frozen LLM. Such an inequitable treatment of vision and language heavily constrains the model's potential. In this paper, we break through this limitation by representing both vision and language in a unified form. Specifically, we introduce a well-designed visual tokenizer to translate the non-linguistic image into a sequence of discrete tokens like a foreign language that LLM can read. The resulting visual tokens encompass high-level semantics worthy of a word and also support dynamic sequence length varying from the image. Coped with this tokenizer, the presented foundation model called LaVIT can handle both image and text indiscriminately under the same generative learning paradigm. This unification empowers LaVIT to serve as an impressive generalist interface to understand and generate multi-modal content simultaneously. Extensive experiments further showcase that it outperforms the existing models by a large margin on massive vision-language tasks. Our code and models are available at https://github.com/jy0205/LaVIT.

cs.CV↗

arXiv2024

Chiral phase transition and spin alignment of vector meson in the Polarized-Polyakov-loop Nambu-Jona-Lasinio model under rotation

By using the extrapolation method, a polarized Polykov-loop potential at finite real angular velocity is constructed from the lattice results at finite imaginary angular velocity. The chiral and deconfinement phase transitions under rotation have been simultaneously investigated in the Polarized-Polyakov-loop Nambu-Jona-Lasinio (PPNJL) model. It is observed that both critical temperatures of deconfinement and chiral phase transition increase with the angular velocity, which is in consistent with lattice results. The spin alignment of vector meson has the negative deviation of $ρ_{00} -1/3$ under rotation, and the deviation in the PPNJL model is much more significant than that in the NJL model and the quark coalescence model, which revealing the important role of rotating gluons on the quark polarization.

hep-ph↗

arXiv2024

Harder Tasks Need More Experts: Dynamic Routing in MoE Models

In this paper, we introduce a novel dynamic expert selection framework for Mixture of Experts (MoE) models, aiming to enhance computational efficiency and model performance by adjusting the number of activated experts based on input difficulty. Unlike traditional MoE approaches that rely on fixed Top-K routing, which activates a predetermined number of experts regardless of the input's complexity, our method dynamically selects experts based on the confidence level in expert selection for each input. This allows for a more efficient utilization of computational resources, activating more experts for complex tasks requiring advanced reasoning and fewer for simpler tasks. Through extensive evaluations, our dynamic routing method demonstrates substantial improvements over conventional Top-2 routing across various benchmarks, achieving an average improvement of 0.7% with less than 90% activated parameters. Further analysis shows our model dispatches more experts to tasks requiring complex reasoning skills, like BBH, confirming its ability to dynamically allocate computational resources in alignment with the input's complexity. Our findings also highlight a variation in the number of experts needed across different layers of the transformer model, offering insights into the potential for designing heterogeneous MoE frameworks. The code and models are available at https://github.com/ZhenweiAn/Dynamic_MoE.

cs.LG↗

arXiv2024

Implicit high-order gas-kinetic schemes for compressible flows on three-dimensional unstructured meshes I: steady flows

In the previous studies, the high-order gas-kinetic schemes (HGKS) have achieved successes for unsteady flows on three-dimensional unstructured meshes. In this paper, to accelerate the rate of convergence for steady flows, the implicit non-compact and compact HGKSs are developed. For non-compact scheme, the simple weighted essentially non-oscillatory (WENO) reconstruction is used to achieve the spatial accuracy, where the stencils for reconstruction contain two levels of neighboring cells. Incorporate with the nonlinear generalized minimal residual (GMRES) method, the implicit non-compact HGKS is developed. In order to improve the resolution and parallelism of non-compact HGKS, the implicit compact HGKS is developed with Hermite WENO (HWENO) reconstruction, in which the reconstruction stencils only contain one level of neighboring cells. The cell averaged conservative variable is also updated with GMRES method. Simultaneously, a simple strategy is used to update the cell averaged gradient by the time evolution of spatial-temporal coupled gas distribution function. To accelerate the computation, the implicit non-compact and compact HGKSs are implemented with the graphics processing unit (GPU) using compute unified device architecture (CUDA). A variety of numerical examples, from the subsonic to supersonic flows, are presented to validate the accuracy, robustness and efficiency of both inviscid and viscous flows.

math.NA↗