Searcharxiv⌕ Search

arXiv subjects

Zhiyuan Li

Publications and source records attributed to Zhiyuan Li.

At least 253 records · Page 14Linked to original sources
arXiv2023

Bayesian optimization with active learning of Ta-Nb-Hf-Zr-Ti system for spin transport properties

Designing materials with enhanced spin charge conversion, i.e., with high spin Hall conductivity (SHC) and low longitudinal electric conductivity (hence large spin Hall angle (SHA)), is a challenging task, especially in the presence of a vast chemical space for compositionally complex alloys (CCAs). In this work, focusing on the Ta-Nb-Hf-Zr-Ti system, we confirm that CCAs exhibit significant spin Hall conductivities and propose a multi-objective Bayesian optimization approach (MOBO) incorporated with active learning (AL) in order to screen for the optimal compositions with significant SHC and SHA. As a result, within less than 5 iterations we are able to target the TaZr-dominated systems displaying both high magnitudes of SHC (~-2.0 (10$^{-3}$ $Ω$ cm)$^{-1}$) and SHA (~0.03). The SHC is mainly ascribed to the extrinsic skew scattering mechanism. Our work provides an efficient route for identifying new materials with significant SHE, which can be straightforwardly generalized to optimize other properties in a vast chemical space.

cond-mat.mtrl-sci↗
arXiv2023

AMUSE-antlia I: Nuclear X-ray properties of early-type galaxies in a dynamically young galaxy cluster

To understand the formation and growth of supermassive black holes (SMBHs) and their co-evolution with host galaxies, it is essential to know the impact of environment on the activity of active galactic nuclei (AGN). We present new Chandra X-ray observations of nuclear emission from member galaxies in the Antlia cluster, the nearest non-cool core and the nearest merging galaxy cluster, residing at D = 35.2 Mpc. Its inner region, centered on two dominant galaxies NGC 3268 and NGC 3258, has been mapped with three deep Chandra ACIS-I pointings. Nuclear X-ray sources are detected in 7/84 (8.3%) early-type galaxies (ETG) and 2/8 (25%) late-type galaxies with a median detection limit of 8x10^38 erg/s. All nuclear X-ray sources but one have a corresponding radio continuum source detected by MeerKAT at the L-band. Nuclear X-ray sources detected in early-type galaxies are considered as the genuine X-ray counterpart of low-luminosity AGN. When restricted to a detection limit of logLx(erg/s) > 38.9 and a stellar mass of 10 < log Ms(Msun) <11.6, 6/11 (54.5%) ETG are found to contain an X-ray AGN in Antlia, exceeding the AGN occupation fraction of 7/39 (18.0%) and 2/12 (16.7%) in the more relaxed, cool core clusters, Virgo and Fornax, respectively, and rivaling that of the AMUSE-Field ETG of 27/49 (55.1%). Furthermore, more than half of the X-ray AGN in Antlia are hosted by its younger subcluster, centered on NGC 3258. We believe that this is because SMBH activity is enhanced in a dynamically young cluster compared to relatively relaxed clusters.

astro-ph.GA↗
arXiv2023

Algebraic reverse Khovanskii--Teissier inequality via Okounkov bodies

Let $X$ be a projective variety of dimension $n$ over an algebraically closed field of arbitrary characteristic and let $A, B, C$ be nef divisors on $X$. We show that for any integer $1\leq k\leq n-1$, $$ (B^k\cdot A^{n-k})\cdot (A^k\cdot C^{n-k})\geq \frac{k!(n-k)!}{n!}(A^n)\cdot (B^k\cdot C^{n-k}). $$ The same inequality in the analytic setting was obtained by Lehmann and Xiao for compact Kähler manifolds using the Calabi--Yau theorem, while our approach is purely algebraic using (multipoint) Okounkov bodies. We also discuss applications of this inequality to Bézout-type inequalities and inequalities on degrees of dominant rational self-maps.

math.AG↗
arXiv2023

Distributionally Robust Optimization and Invariant Representation Learning for Addressing Subgroup Underrepresentation: Mechanisms and Limitations

Spurious correlation caused by subgroup underrepresentation has received increasing attention as a source of bias that can be perpetuated by deep neural networks (DNNs). Distributionally robust optimization has shown success in addressing this bias, although the underlying working mechanism mostly relies on upweighting under-performing samples as surrogates for those underrepresented in data. At the same time, while invariant representation learning has been a powerful choice for removing nuisance-sensitive features, it has been little considered in settings where spurious correlations are caused by significant underrepresentation of subgroups. In this paper, we take the first step to better understand and improve the mechanisms for debiasing spurious correlation due to subgroup underrepresentation in medical image classification. Through a comprehensive evaluation study, we first show that 1) generalized reweighting of under-performing samples can be problematic when bias is not the only cause for poor performance, while 2) naive invariant representation learning suffers from spurious correlations itself. We then present a novel approach that leverages robust optimization to facilitate the learning of invariant representations at the presence of spurious correlations. Finetuned classifiers utilizing such representation demonstrated improved abilities to reduce subgroup performance disparity, while maintaining high average and worst-group performance.

cs.CV↗
arXiv2023

Sharpness Minimization Algorithms Do Not Only Minimize Sharpness To Achieve Better Generalization

Despite extensive studies, the underlying reason as to why overparameterized neural networks can generalize remains elusive. Existing theory shows that common stochastic optimizers prefer flatter minimizers of the training loss, and thus a natural potential explanation is that flatness implies generalization. This work critically examines this explanation. Through theoretical and empirical investigation, we identify the following three scenarios for two-layer ReLU networks: (1) flatness provably implies generalization; (2) there exist non-generalizing flattest models and sharpness minimization algorithms fail to generalize, and (3) perhaps most surprisingly, there exist non-generalizing flattest models, but sharpness minimization algorithms still generalize. Our results suggest that the relationship between sharpness and generalization subtly depends on the data distributions and the model architectures and sharpness minimization algorithms do not only minimize sharpness to achieve better generalization. This calls for the search for other explanations for the generalization of over-parameterized neural networks.

cs.LG↗
arXiv2023

Prototype-Driven and Multi-Expert Integrated Multi-Modal MR Brain Tumor Image Segmentation

For multi-modal magnetic resonance (MR) brain tumor image segmentation, current methods usually directly extract the discriminative features from input images for tumor sub-region category determination and localization. However, the impact of information aliasing caused by the mutual inclusion of tumor sub-regions is often ignored. Moreover, existing methods usually do not take tailored efforts to highlight the single tumor sub-region features. To this end, a multi-modal MR brain tumor segmentation method with tumor prototype-driven and multi-expert integration is proposed. It could highlight the features of each tumor sub-region under the guidance of tumor prototypes. Specifically, to obtain the prototypes with complete information, we propose a mutual transmission mechanism to transfer different modal features to each other to address the issues raised by insufficient information on single-modal features. Furthermore, we devise a prototype-driven feature representation and fusion method with the learned prototypes, which implants the prototypes into tumor features and generates corresponding activation maps. With the activation maps, the sub-region features consistent with the prototype category can be highlighted. A key information enhancement and fusion strategy with multi-expert integration is designed to further improve the segmentation performance. The strategy can integrate the features from different layers of the extra feature extraction network and the features highlighted by the prototypes. Experimental results on three competition brain tumor segmentation datasets prove the superiority of the proposed method.

eess.IV↗
arXiv2023

Bond-based nonlocal models by nonlocal operator method in symmetric support domain

This paper is concerned with the energy decomposition of various nonlocal models, including elasticity, thin plates, and gradient elasticity, to arrive at bond-based nonlocal models in which the bond force depends only on the deformation of a single bond. By assuming an appropriate form of bond force and using energy equivalence between local and nonlocal models, several very concise bond-based models are derived. We also revisit the nonlocal operator methods and study the simplified form of second-order NOM in the symmetric support domain. A bent-bond consisting of three points is proposed to describe the curvature and moment. To model the damage, a rule based on Griffith theory for the critical normal strain of the bond is proposed in analogy to the phase field model, which can be applied individually to each bond and provides strain localization. With this rule, the crack direction can be automatically predicted by simply cutting the bond, giving comparable results to the phase field method. At the same time, a damage rule for critical shear strains in shear fractures is proposed. Furthermore, an incremental form of the plasticity model for bond reaction force is derived. Several numerical examples are presented to further validate the nonlocal bond-based models.

physics.class-ph↗
arXiv2023

The Inductive Bias of Flatness Regularization for Deep Matrix Factorization

Recent works on over-parameterized neural networks have shown that the stochasticity in optimizers has the implicit regularization effect of minimizing the sharpness of the loss function (in particular, the trace of its Hessian) over the family zero-loss solutions. More explicit forms of flatness regularization also empirically improve the generalization performance. However, it remains unclear why and when flatness regularization leads to better generalization. This work takes the first step toward understanding the inductive bias of the minimum trace of the Hessian solutions in an important setting: learning deep linear networks from linear measurements, also known as \emph{deep matrix factorization}. We show that for all depth greater than one, with the standard Restricted Isometry Property (RIP) on the measurements, minimizing the trace of Hessian is approximately equivalent to minimizing the Schatten 1-norm of the corresponding end-to-end matrix parameters (i.e., the product of all layer matrices), which in turn leads to better generalization. We empirically verify our theoretical findings on synthetic datasets.

cs.LG↗
arXiv2023

Searching for Radio Outflows from M31* with VLBI Observations

As one of the nearest and most dormant supermassive black holes (SMBHs), M31* provides a rare but promising opportunity for studying the physics of black hole accretion and feedback at the quiescent state. Previous Karl G. Jansky Very Large Array (VLA) observations with an arcsec resolution have detected M31* as a compact radio source over centimeter wavelengths, but the steep radio spectrum suggests optically-thin synchrotron radiation from an outflow driven by a hot accretion flow onto the SMBH. Aiming to probe the putative radio outflow, we have conducted milli-arcsec-resolution very long baseline interferometric (VLBI) observations of M31* in 2016, primarily at 5 GHz and combining the Very Long Baseline Array, Tianma-65m and Shanghai-25m Radio Telescopes. Despite the unprecedented simultaneous resolution and sensitivity achieved, no significant ($\gtrsim 3σ$) signal is detected at the putative position of M31* given an RMS level of $\rm 5.9~μJy\ beam^{-1}$, thus ruling out a point-like source with a peak flux density comparable to that ($\sim30~μJy\ beam^{-1}$) measured by the VLA observations taken in 2012. We disfavor the possibility that M31* has substantially faded since 2012, in view that a 2017 VLA observation successfully detected M31* at a historically-high peak flux density ($\sim75~μJy\ beam^{-1}$ at 6 GHz). Instead, the non-detection of the VLBI observations is best interpreted as the arcsec-scale core being resolved out at the milli-arcsec-scale, suggesting an intrinsic size of M31* at 5 GHz larger than $\sim300$ times the Schwarzschild radius. Such extended radio emission may originate from a hot wind driven by the weakly accreting SMBH.

astro-ph.GA↗
arXiv2023

Learning Empirical Bregman Divergence for Uncertain Distance Representation

Deep metric learning techniques have been used for visual representation in various supervised and unsupervised learning tasks through learning embeddings of samples with deep networks. However, classic approaches, which employ a fixed distance metric as a similarity function between two embeddings, may lead to suboptimal performance for capturing the complex data distribution. The Bregman divergence generalizes measures of various distance metrics and arises throughout many fields of deep metric learning. In this paper, we first show how deep metric learning loss can arise from the Bregman divergence. We then introduce a novel method for learning empirical Bregman divergence directly from data based on parameterizing the convex function underlying the Bregman divergence with a deep learning setting. We further experimentally show that our approach performs effectively on five popular public datasets compared to other SOTA deep metric learning methods, particularly for pattern recognition problems.

cs.CV↗
arXiv2023

Performing SU($d$) operations and rudimentary algorithms in a superconducting transmon qudit for $d=3$ and $d=4$

Quantum computation architecture based on $d$-level systems, or qudits, has attracted considerable attention recently due to their enlarged Hilbert space. Extensive theoretical and experimental studies have addressed aspects of algorithms and benchmarking techniques for qudit-based quantum computation and quantum information processing. Here, we report a physical realization of qudit with upto 4 embedded levels in a superconducting transmon, demonstrating high-fidelity initialization, manipulation, and simultaneous multi-level readout. In addition to constructing SU($d$) operations and benchmarking protocols for quantum state tomography, quantum process tomography, and randomized benchmarking etc, we experimentally carry out these operations for $d=3$ and $d=4$. Moreover, we perform prototypical quantum algorithms and observe outcomes consistent with expectations. Our work will hopefully stimulate further research interest in developing manipulation protocols and efficient applications for quantum processors with qudits.

quant-ph↗
arXiv2023

Hydrodynamic Evolution of Sgr A East: The Imprint of A Supernova Remnant in the Galactic Center

We perform three-dimensional numerical simulations to study the hydrodynamic evolution of Sgr A East, the only known supernova remnant (SNR) in the center of our Galaxy, to infer its debated progenitor SN type and its potential impact on the Galactic center environment. Three sets of simulations are performed, each of which represents a represent a certain type of SN explosion (SN Iax, SN Ia or core-collapse SN) expanding against a nuclear outflow of hot gas driven by massive stars, whose thermodynamical properties have been well established by previous work and fixed in the simulations. All three simulations can simultaneously roughly reproduce the extent of Sgr A East and the position and morphology of an arc-shaped thermal X-ray feature, known as the "ridge". Confirming previous work, our simulations show that the ridge is the manifestation of a strong collision between the expanding SN ejecta and the nuclear outflow. The simulation of the core-collapse SN, with an assumed explosion energy of 5x10^50 erg and an ejecta mass of 10 M_sun, can well match the X-ray flux of the ridge, whereas the simulations of the SN Iax and SN Ia explosions underpredict its X-ray emission, due to a smaller ejecta mass. All three simulations constrain the age of Sgr A East to be <1500 yr and predict that the ridge should fade out over the next few hundred years. We address the implications of these results for our understanding of the Galactic center environment.

astro-ph.HE↗
arXiv2023

Learning Generalized Hybrid Proximity Representation for Image Recognition

Recently, deep metric learning techniques received attention, as the learned distance representations are useful to capture the similarity relationship among samples and further improve the performance of various of supervised or unsupervised learning tasks. We propose a novel supervised metric learning method that can learn the distance metrics in both geometric and probabilistic space for image recognition. In contrast to the previous metric learning methods which usually focus on learning the distance metrics in Euclidean space, our proposed method is able to learn better distance representation in a hybrid approach. To achieve this, we proposed a Generalized Hybrid Metric Loss (GHM-Loss) to learn the general hybrid proximity features from the image data by controlling the trade-off between geometric proximity and probabilistic proximity. To evaluate the effectiveness of our method, we first provide theoretical derivations and proofs of the proposed loss function, then we perform extensive experiments on two public datasets to show the advantage of our method compared to other state-of-the-art metric learning methods.

cs.CV↗
arXiv2023

A Diverse Population of z ~ 2 ULIRGs Revealed by JWST Imaging

Four ultra-luminous infrared galaxies (ULIRGs) observed with JWST/NIRcam in the Cosmos Evolution Early Release Science program offer an unbiased preview of the $z\approx2$ ULIRG population. The objects were originally selected at 24 $μ$m and have strong polycyclic aromatic hydrocarbon emission features observed with Spitzer/IRS. The four objects have similar stellar masses of ${\sim}10^{11}$ M$_\odot$ but otherwise are quite diverse. One is an isolated disk galaxy, but it has an active nucleus as shown by X-ray observations and by a bright point-source nucleus. Two others are merging pairs with mass ratios of 6-7:1. One has active nuclei in both components, while the other has only one active nucleus: the one in the less-massive neighbor, not the ULIRG. The fourth object is clumpy and irregular and is probably a merger, but there is no sign of an active nucleus. The intrinsic spectral energy distributions for the four AGNs in these systems are typical of type-2 QSOs. This study is consistent with the idea that even if internal processes can produce large luminosities at $z\sim2$, galaxy merging may still be necessary for the most luminous objects. The diversity of these four initial examples suggests that large samples will be needed to understand the $z\approx2$ ULIRG population.

astro-ph.GA↗
arXiv2023

Chandra X-ray Measurement of Gas-phase Heavy Element Abundances in the Central Parsec of the Galaxy

Elemental abundances are key to our understanding of star formation and evolution in the Galactic center. Previous work on this topic has been based on infrared (IR) observations, but X-ray observations have the potential of constraining the abundance of heavy elements, mainly through their K-shell emission lines. Using 5.7 Ms Chandra observations, we provide the first abundance measurement of Si, S, Ar, Ca and Fe, in four prominent diffuse X-ray features located in the central parsec of the Galaxy, which are the manifestation of shock-heated hot gas. A two-temperature, non-equilibrium ionization spectral model is employed to derive the abundances of these five elements. In this procedure, a degeneracy is introduced due to uncertainties in the composition of light elements, in particular, H, C and N. Assuming that the hot gas is H-depleted but C- and N-enriched, as would be expected for a standard scenario in which the hot gas is dominated by Wolf-Rayet star winds, the spectral fit finds a generally subsolar abundance for the heavy elements. If, instead, the light elements had a solar-like abundance, the heavy elements have a fitted abundance of $\sim$1--2 solar. The $α$/Fe abundance ratio, on the other hand, is mostly supersolar and insensitive to the exact composition of the light elements. These results are robust against potential biases due to either a moderate spectral S/N or the presence of non-thermal components. Implications of the measured abundances for the Galactic center environment are addressed.

astro-ph.GA↗
arXiv2023

Initial-boundary value problems for coupled systems of time-fractional diffusion equations

This article deals with the initial-boundary value problem for a moderately coupled system of time-fractional diffusion equations. Defining the mild solution, we establish fundamental unique existence, limited smoothing property and long-time asymptotic behavior of the solution, which mostly inherit those of a single equation. Owing to the coupling effect, we also obtain the uniqueness for an inverse problem on determining all the fractional orders by the single point observation of a single component of the solution.

math.AP↗
arXiv2023

Periodic X-ray sources in the Massive Globular Cluster 47 Tucanae: Evidence for Dynamically Formed Cataclysmic Variables

We present a systematic study of periodic X-ray sources in the massive globular cluster 47 Tuc, utilizing deep archival Chandra observations that resolve the cluster core and recently available eROSITA observations that cover the cluster outskirt. By applying the Gregory-Loredo algorithm, we detect 20 periodic signals among 18 X-ray sources, ranging between 205-95731 second. Fourteen periods are newly discovered in the X-ray band. We classify these periodic sources into four quiescent low-mass X-ray binaries, one milli-second pulsar, two coronally-active binaries and eleven cataclysmic variables (CVs), based on their X-ray temporal and spectral properties, as well as multi-band information. Despite a small sample subject to potential selection bias against faint and non-magnetic CVs, the 11 CVs together define an orbital period distribution significantly different from that of the CVs previously found in the solar neighborhood and the Galactic bulge. In particular, there exists in 47 Tuc an apparent paucity of short-period CVs below the period gap, which might be attributed to a high occupation fraction of non-magnetic CVs. Also characteristic of the 47 Tuc CVs are an overabundance of long-period CVs with a subgiant donor, a substantial fraction of CVs within the period gap, and a steep radial surface density profile. These are best understood as a group of CVs having recently formed via dynamical interactions in the dense cluster core. Despite sufficient sensitivity of the X-ray data, only one periodic source is found between one-third of the half-light radius and the tidal radius, the nature of which is unclear.

astro-ph.HE↗
arXiv2023

A Chandra X-ray Survey of Optically Selected Close Galaxy Pairs: Unexpectedly Low Occupation of Active Galactic Nuclei

High-resolution X-ray observations offer a unique tool for probing the still elusive connection between galaxy mergers and active galactic nuclei (AGNs). We present an analysis of nuclear X-ray emission in an optically selected sample of 92 close galaxy pairs (with projected separations $\lesssim 20$ kpc and line-of-sight velocity offsets $<$ 500 km s$^{-1}$) at low redshift ($\bar{z} \sim 0.07$), based on archival Chandra observations. The parent sample of galaxy pairs is constructed without imposing an optical classification of nuclear activity, thus is largely free of selection effect for or against the presence of an AGN. Nor is this sample biased for or against gas-rich mergers. An X-ray source is detected in 70 of the 184 nuclei, giving a detection rate of $38\%^{+5\%}_{-5\%}$, down to a 0.5-8 keV limiting luminosity of $\lesssim 10^{40}\rm~erg~s^{-1}$. The detected and undetected nuclei show no systematic difference in their host galaxy properties such as galaxy morphology, stellar mass and stellar velocity dispersion. When potential contamination from star formation is avoided (i.e., $L_{\rm 2-10~keV} > 10^{41}\rm~erg~s^{-1}$), the detection rate becomes $18\%^{+3\%}_{-3\%}$ (32/184), which shows no excess compared to the X-ray detection rate of a comparison sample of optically classified single AGNs. The fraction of pairs containing dual AGN is only $2\%^{+2\%}_{-2\%}$. Moreover, most nuclei at the smallest projected separations probed by our sample (a few kpc) have an unexpectedly low apparent X-ray luminosity and Eddington ratio, which cannot be solely explained by circumnuclear obscuration. These findings suggest that close galaxy interaction is not a sufficient condition for triggering a high level of AGN activity.

astro-ph.GA↗