arXiv subjects

Jun Xu

Publications and source records attributed to Jun Xu.

At least 379 records · Page 21Linked to original sources

arXiv2020

Centrality fluctuations and decorrelations in heavy-ion collisions

The centrality or the number of initial-state sources $V$ of the system produced in heavy ion collision is a concept that is not uniquely defined and subject to significant theoretical and experimental uncertainties. We argue that a more robust connection between the initial-state sources with final-state multiplicity could be established from the event-by-event multiplicity correlation between two subevents separated in pseudorapidity, $N_a$ vs $N_b$. This correlation is sensitive to two main types of centrality fluctuations (CF): 1) particle production for each source $p(n)$ which smears the relation between $V$ and $N_a$ used for experimental centrality, and 2) decorrelations between the sources in the two subevents $V_b$ and $V_a$. The CF is analyzed in terms of cumulants of $V_b$ and $N_b$ as a function of $N_a$, i.e. experimental centrality is defined with $N_a$. We found that the mean values $\langle V_b\rangle_{N_a}$ and $\langle N_b\rangle_{N_a}$ increase linearly with $N_a$ in mid-central collisions, but flatten out in ultra-central collisions. Such non-linear behavior is sensitive to the centrality resolution of $N_a$. In the presence of centrality decorrelations, the scaled variances $\langle(δV_b)^2\rangle/\langle V_b\rangle$ and $\langle(δN_b)^2\rangle/\langle N_b\rangle$ are found to decrease linearly with $N_a$ in mid-central collisions, while the $p(n)$ leads to another sharp decrease in the ultra-central region. The higher-order cumulants of $V_b$ and $N_b$ show interesting but rather complex behaviors which deserve further studies. Our results suggest that one can use the cumulants of the two-dimensional multiplicity correlation, especially the mean and variance, to constrain the particle production mechanism as well as the longitudinal fluctuations of the initial-state sources.

nucl-th↗

arXiv2019

Hadronization using the Wigner function approach for a multiphase transport model

In the string melting version of a multiphase transport model, the hadronization algorithm has been improved by favoring parton combinations close in not only coordinate space but also momentum space. Formation probabilities of mesons, baryons, and antibaryons during hadronization are determined by their corresponding Wigner functions for the valence parton combinations with no free parameters, and the net baryon, electric, and strangeness charges are conserved during quark coalescence. Effects of the hadronization on anisotropic flows, collision energy dependence of the proton directed flow, relative particle yield ratios, as well as di-hadron correlations in relativistic heavy-ion collisions at the energies ranging from RHIC beam energy scan to LHC have been extensively discussed.

nucl-th↗

arXiv2019

Efficient hinging hyperplanes neural network and its application in nonlinear system identification

In this paper, the efficient hinging hyperplanes (EHH) neural network is proposed based on the model of hinging hyperplanes (HH). The EHH neural network is a distributed representation, the training of which involves solving several convex optimization problems and is fast. It is proved that for every EHH neural network, there is an equivalent adaptive hinging hyperplanes (AHH) tree, which was also proposed based on the model of HH and find good applications in system identification. The construction of the EHH neural network includes 2 stages. First the initial structure of the EHH neural network is randomly determined and the Lasso regression is used to choose the appropriate network. To alleviate the impact of randomness, secondly, the stacking strategy is employed to formulate a more general network structure. Different from other neural networks, the EHH neural network has interpretability ability, which can be easily obtained through its ANOVA decomposition (or interaction matrix). The interpretability can then be used as a suggestion for input variable selection. The EHH neural network is applied in nonlinear system identification, the simulation results show that the regression vector selected is reasonable and the identification speed is fast, while at the same time, the simulation accuracy is satisfactory.

eess.SY↗

arXiv2019

Nucleon spin polarization in intermediate-energy heavy-ion collisions

Based on a spin-dependent Boltzmann-Uehling-Uhlenbeck transport model with spin-dependent potentials incorporated using the lattice Hamiltonian method, we have studied the global spin polarization perpendicular to the reaction plane as well as the local spin polarization in the longitudinal direction in non-central intermediate-energy heavy-ion collisions, as an extension of similar studies in relativistic heavy-ion collisions. Both the global and the local spin polarizations are found to be mostly dominated by the time-odd component of the nuclear spin-orbit potential. Impacts of various theoretical uncertainties on the nucleon spin polarization as well as its dependence on the beam energy and impact parameter are discussed. Our study serves as a baseline for understanding the spin polarization mechanism of strongly interacting matter produced in heavy-ion collisions dominated by nucleon degree of freedom.

nucl-th↗

arXiv2019

Precision measurements of the gradient of the Casimir force between ultra clean metallic surfaces at larger separations

We report precision measurements of the Casimir interaction at larger separation distances between the Au-coated surfaces of a sphere and a plate in ultrahigh vacuum using a much softer cantilever of the dynamic atomic force microscope-based setup and two-step cleaning procedure of the vacuum chamber and test body surfaces by means of UV light and Ar-ion bombardment. Compared to the previously performed experiment, two more measurement sets for the gradient of the Casimir force are provided which confirmed and slightly improved the results. Next, additional measurements have been performed with a factor of two larger oscillation amplitude of the cantilever. This allowed obtaining meaningful results at much larger separation distances. The comparison of the measurement data with theoretical predictions of the Lifshitz theory using the dissipative Drude model to describe the response of Au to the low-frequency electromagnetic field fluctuations shows that this theoretical approach is experimentally excluded over the distances from 250 to 1100nm (i.e., a major step forward has been made as compared to the previous work where it was excluded up to only 820nm). The theoretical approach using the dissipationless plasma model at low frequencies is shown to be consistent with the data over the entire measurement range from 250 to 1300nm. The possibilities to explain these puzzling results are discussed.

quant-ph↗

arXiv2019

Spectro-polarimetric Observations at the NVST: I. Instrumental Polarization Calibration and Primary Measurements

This paper is devoted to the primary spectro-polarimetric observation performed at the New Vacuum Solar Telescope of China since 2017, and our aim is to precisely evaluate the real polarimetric accuracy and sensitivity of this polarimetry by using full Stokes spectro-polarimetric observations of the photospheric line Fe I 532.4 nm. In the work, we briefly describe the salient characteristic of the NVST as a polarimeter in technology and then characterize its instrumental polarization based on the operation in 2017 and 2019. It is verified that the calibration method making use of the instrumental polarization calibration unit (ICU) is stable and credible. The calibration accuracy can reach up to 3$\times 10^{-3}$ . Based on the scientific observation of the NOAA 12645 on April 5th, 2017, we estimate that the residual cross-talk from Stokes $I$ to Stokes $Q$, $U$ and $V$, after the instrumental polarization calibration, is about 4$\times10^{-3}$ on average, which is consistent with the calibration accuracy and close to the photon noise. The polarimetric sensitivity (i.e., the detection limit) for polarized light is of the order of $10^{-3}$ with an integration time over 20 seconds. Slow modulation rate is indeed an issue for the present system. The present NVST polarimeter is expected to be integrated with an high-order adaptive optics system and a field scanner to realize 2D magnetic field vector measurements in the following instrumentation update.

astro-ph.SR↗

arXiv2019

Isospin splitting of pion elliptic flow in relativistic heavy-ion collisions

Based on the framework of an extended multiphase transport model with mean-field potentials in both the partonic phase and the hadronic phase, we explain the elliptic flow difference between $π^+$ and $π^-$ in the beam-energy scan program at the relativistic heavy-ion collider by incorporating the vector-isovector potential for quarks and antiquarks with different isospins. It is found that the isospin splitting of pion elliptic flow favors a strong vector-isovector interaction, and thus serves as a probe of the quark matter equation of state as well as the QCD phase structure at finite baryon and isospin chemical potentials.

nucl-th↗

arXiv2019

Comparison of heavy-ion transport simulations: Collision integral with pions and $Δ$ resonances in a box

We compare ten transport codes for a system confined in a box, aiming at improved handling of the production of $Δ$ resonances and pions, which is indispensable for constraining high-density symmetry energy from observables such as the $π^-/π^+$ yield ratio in heavy-ion collisions. The system in a box is initialized with nucleons at saturation density and at 60 MeV temperature. The reactions $NN\leftrightarrow NΔ$ and $Δ\leftrightarrow Nπ$ are implemented, but the Pauli blocking and the mean-field potential are deactivated in the present comparison. Results are compared to those from the two reference cases of a chemically equilibrated ideal gas mixture and of the rate equation. In the results of the numbers of $Δ$ and $π$, deviations from the reference values are observed in many codes, and they depend significantly on the size of the time step. These deviations are tied to different ways in ordering the sequence of collisions and decays, that take place in the same time step. Better agreements are seen in the reaction rates and the number ratios among the isospin species of $Δ$ and $π$. These are, however, affected by the correlations, which are absent in the Boltzmann equation, but are induced by the way particle scatterings are treated in transport calculations. The uncertainty in the transport-code predictions of the $π^-/π^+$ ratio for the system initialized at n/p = 1.5, after letting the existing $Δ$ resonances decay, is found to be within a few percent, which is sufficiently small so that it does not strongly impact constraining the high-density symmetry energy from heavy-ion collisions. Most of the sources of uncertainties have been understood, and individual codes may be further improved. This investigation will be extended in the future to heavy-ion collisions to ensure the problems identified here remain under control.

nucl-th↗

arXiv2019

Scaled Simplex Representation for Subspace Clustering

The self-expressive property of data points, i.e., each data point can be linearly represented by the other data points in the same subspace, has proven effective in leading subspace clustering methods. Most self-expressive methods usually construct a feasible affinity matrix from a coefficient matrix, obtained by solving an optimization problem. However, the negative entries in the coefficient matrix are forced to be positive when constructing the affinity matrix via exponentiation, absolute symmetrization, or squaring operations. This consequently damages the inherent correlations among the data. Besides, the affine constraint used in these methods is not flexible enough for practical applications. To overcome these problems, in this paper, we introduce a scaled simplex representation (SSR) for subspace clustering problem. Specifically, the non-negative constraint is used to make the coefficient matrix physically meaningful, and the coefficient vector is constrained to be summed up to a scalar s<1 to make it more discriminative. The proposed SSR based subspace clustering (SSRSC) model is reformulated as a linear equality-constrained problem, which is solved efficiently under the alternating direction method of multipliers framework. Experiments on benchmark datasets demonstrate that the proposed SSRSC algorithm is very efficient and outperforms state-of-the-art subspace clustering methods on accuracy. The code can be found at https://github.com/csjunxu/SSRSC.

cs.CV↗

arXiv2019

Simulating chiral anomalies with spin dynamics

Considering that the chiral kinetic equations of motion (CEOM) can be derived from the spin kinetic equations of motion (SEOM) for massless particles with approximations, we simulate the chiral anomalies by using the latter in a box system with the periodic boundary condition under a uniform external magnetic field. We found that the chiral magnetic effect is weaker while the damping of the chiral magnetic wave is stronger from the SEOM compared with that from the CEOM. In addition, effects induced by chiral anomalies from the SEOM are less sensitive to the decay of the magnetic field than from the CEOM due to the spin relaxation process.

nucl-th↗

arXiv2019

RANet: Ranking Attention Network for Fast Video Object Segmentation

Despite online learning (OL) techniques have boosted the performance of semi-supervised video object segmentation (VOS) methods, the huge time costs of OL greatly restrict their practicality. Matching based and propagation based methods run at a faster speed by avoiding OL techniques. However, they are limited by sub-optimal accuracy, due to mismatching and drifting problems. In this paper, we develop a real-time yet very accurate Ranking Attention Network (RANet) for VOS. Specifically, to integrate the insights of matching based and propagation based methods, we employ an encoder-decoder framework to learn pixel-level similarity and segmentation in an end-to-end manner. To better utilize the similarity maps, we propose a novel ranking attention module, which automatically ranks and selects these maps for fine-grained VOS performance. Experiments on DAVIS-16 and DAVIS-17 datasets show that our RANet achieves the best speed-accuracy trade-off, e.g., with 33 milliseconds per frame and J&F=85.5% on DAVIS-16. With OL, our RANet reaches J&F=87.1% on DAVIS-16, exceeding state-of-the-art VOS methods. The code can be found at https://github.com/Storife/RANet.

cs.CV↗

arXiv2019

Power Minimization for Wireless Backhaul Based Ultra-Dense Cache-enabled C-RAN

This correspondence paper investigates joint design of small base station (SBS) clustering, multicast beamforming for access and backhaul links, as well as frequency allocation in backhaul transmission to minimize the total power consumption for wireless backhaul based ultra-dense cache-enabled cloud radio access network (C-RAN). To solve this nontrivial problem, we develop a low-complexity algorithm, which is a combination of smoothed ${\ell _0}{\text{-norm}}$ approximation and convex-concave procedure. Simulation results show that the proposed algorithm converges fast and greatly reduces the backhaul traffic.

eess.SP↗

arXiv2019

HorNet: A Hierarchical Offshoot Recurrent Network for Improving Person Re-ID via Image Captioning

Person re-identification (re-ID) aims to recognize a person-of-interest across different cameras with notable appearance variance. Existing research works focused on the capability and robustness of visual representation. In this paper, instead, we propose a novel hierarchical offshoot recurrent network (HorNet) for improving person re-ID via image captioning. Image captions are semantically richer and more consistent than visual attributes, which could significantly alleviate the variance. We use the similarity preserving generative adversarial network (SPGAN) and an image captioner to fulfill domain transfer and language descriptions generation. Then the proposed HorNet can learn the visual and language representation from both the images and captions jointly, and thus enhance the performance of person re-ID. Extensive experiments are conducted on several benchmark datasets with or without image captions, i.e., CUHK03, Market-1501, and Duke-MTMC, demonstrating the superiority of the proposed method. Our method can generate and extract meaningful image captions while achieving state-of-the-art performance.

cs.CV↗

arXiv2019

Semi-Supervised Self-Growing Generative Adversarial Networks for Image Recognition

Image recognition is an important topic in computer vision and image processing, and has been mainly addressed by supervised deep learning methods, which need a large set of labeled images to achieve promising performance. However, in most cases, labeled data are expensive or even impossible to obtain, while unlabeled data are readily available from numerous free on-line resources and have been exploited to improve the performance of deep neural networks. To better exploit the power of unlabeled data for image recognition, in this paper, we propose a semi-supervised and generative approach, namely the semi-supervised self-growing generative adversarial network (SGGAN). Label inference is a key step for the success of semi-supervised learning approaches. There are two main problems in label inference: how to measure the confidence of the unlabeled data and how to generalize the classifier. We address these two problems via the generative framework and a novel convolution-block-transformation technique, respectively. To stabilize and speed up the training process of SGGAN, we employ the metric Maximum Mean Discrepancy as the feature matching objective function and achieve larger gain than the standard semi-supervised GANs (SSGANs), narrowing the gap to the supervised methods. Experiments on several benchmark datasets show the effectiveness of the proposed SGGAN on image recognition and facial attribute recognition tasks. By using the training data with only 4% labeled facial attributes, the SGGAN approach can achieve comparable accuracy with leading supervised deep learning methods with all labeled facial attributes.

cs.CV↗

arXiv2019

Nuclear matter properties at finite temperatures from effective interactions

We study if commonly used nucleon-nucleon effective interactions, obtained from fitting the properties of cold nuclear matter and of finite nuclei, can properly describe the hot dense nuclear matter produced in intermediate-energy heavy-ion collisions. We use two representative effective interactions, i.e., an improved isospin- and momentum-dependent interaction with its isovector part calibrated by the results from the \emph{ab initio} non-perturbative self-consistent Green's function (SCGF) approach with chiral forces, and a Skyme-type interaction fitted to the equation of state of cold nuclear matter from chiral effective many-body perturbation theory and the binding energy of finite nuclei. In the mean-field approximation, we evaluate the equation of state and the single-nucleon potential for nuclear matter at finite temperatures and compare them to those from the SCGF approach. We find that the improved isospin- and momentum-dependent interaction reproduces reasonably well the SCGF results due to its weaker momentum dependence of the mean-field potential than in the Skyrme-type interaction. Our study thus indicates that effective interactions with the correct momentum dependence of the mean-filed potential can properly describe the properties of hot dense nuclear matter and are thus suitable for use in transport models to study heavy-ion collisions at intermediate energies.

nucl-th↗

arXiv2019

Examining the Casimir puzzle with upgraded technique and advanced surface cleaning

We performed measurements of the gradient of the Casimir force between Au-coated surfaces of a sphere and a plate by means of significantly upgraded dynamic atomic force microscope (AFM) based technique. By introducing combined cleaning procedure of interior surfaces of the vacuum chamber and the test bodies by means of UV light and Ar ions, we reached higher vacuum and eliminated the role of electrostatic patches. Furthermore, the use of much softer cantilever allowed a sixfold decrease of the systematic error in measuring the force gradient. The experimental data are compared with theoretical predictions of the Lifshitz theory taking into account corrections due to the inaccuracy of the proximity force approximation and that due to surface roughness. It is shown that the theoretical approach accounting for the relaxation properties of free electrons is excluded by the data up to a larger than previous sphere-plate separation of 820~nm, whereas an alternative approach is found in a very good agreement with the data. Importance of these results in connection with the foundations of quantum statistical physics is discussed.

cond-mat.mes-hall↗

arXiv2019

DeepRank: A New Deep Architecture for Relevance Ranking in Information Retrieval

This paper concerns a deep learning approach to relevance ranking in information retrieval (IR). Existing deep IR models such as DSSM and CDSSM directly apply neural networks to generate ranking scores, without explicit understandings of the relevance. According to the human judgement process, a relevance label is generated by the following three steps: 1) relevant locations are detected, 2) local relevances are determined, 3) local relevances are aggregated to output the relevance label. In this paper we propose a new deep learning architecture, namely DeepRank, to simulate the above human judgment process. Firstly, a detection strategy is designed to extract the relevant contexts. Then, a measure network is applied to determine the local relevances by utilizing a convolutional neural network (CNN) or two-dimensional gated recurrent units (2D-GRU). Finally, an aggregation network with sequential integration and term gating mechanism is used to produce a global relevance score. DeepRank well captures important IR characteristics, including exact/semantic matching signals, proximity heuristics, query term importance, and diverse relevance requirement. Experiments on both benchmark LETOR dataset and a large scale clickthrough data show that DeepRank can significantly outperform learning to ranking methods, and existing deep learning methods.

cs.IR↗

arXiv2019

Neural or Statistical: An Empirical Study on Language Models for Chinese Input Recommendation on Mobile

Chinese input recommendation plays an important role in alleviating human cost in typing Chinese words, especially in the scenario of mobile applications. The fundamental problem is to predict the conditional probability of the next word given the sequence of previous words. Therefore, statistical language models, i.e.~n-grams based models, have been extensively used on this task in real application. However, the characteristics of extremely different typing behaviors usually lead to serious sparsity problem, even n-gram with smoothing will fail. A reasonable approach to tackle this problem is to use the recently proposed neural models, such as probabilistic neural language model, recurrent neural network and word2vec. They can leverage more semantically similar words for estimating the probability. However, there is no conclusion on which approach of the two will work better in real application. In this paper, we conduct an extensive empirical study to show the differences between statistical and neural language models. The experimental results show that the two different approach have individual advantages, and a hybrid approach will bring a significant improvement.

cs.CL↗