arXiv subjects

Xintian Wu

Publications and source records attributed to Xintian Wu.

18 recordsLinked to original sources

arXiv2025

The exact relation between the entanglement entropies of the $XY$ and quantum Ising chains with free and fixed boundary conditions

The entanglement entropies of $XY$ chains and quantum Ising chains (QICs) with fixed boundary conditions are studied here. Three kinds of boundary conditions (BCs) are considered: fixed up--up or down--down (the spins at both ends are aligned in the same direction), fixed up--down or down--up (the spins at the two ends are aligned in opposite directions), and fixed--free (the spin at one end is aligned, and the other end is free). It is shown that i) the entanglement entropy of an $XY$ chain with a fixed--free BC is the sum of those of QICs with a fixed--free BC and with a free--free BC; ii) the entanglement entropy of an $XY$ chain with a fixed up--up boundary condition is the sum of those of QICs with a fixed up--up BC and with a free--free BC; and iii) the entanglement entropy of an $XY$ chain with a fixed up--down BC is the sum of that of a QIC with a fixed up--up BC and that of the first excited state of a QIC with a free--free BC.

cond-mat.stat-mech↗

arXiv2023

First order phase transition in the few-body XY-models with surface fields

We investigate the one-dimensional finite-size XY model with opposing surface fields in the X direction. Exact solutions are obtained for the two-site and three-site models, while numerical methods are employed for models with more than three sites. Remarkably, first-order quantum phase transitions are observed in this system. At the phase transition point, the energy gap closes linearly, and the magnetization at each site undergoes a discontinuous jump. Additionally, we identify a $Z_2$ symmetry that accompanies the phase transition and its associated symmetry change. Notably, the first-order phase transition in finite-size systems does not exhibit the conventional finite-size rounding effect. On the contrary, there exists a counterintuitive finite-size effect: the amplitude of the jump in magnetization at each site decreases as the lattice size increases. Interestingly, lattices with an even number of sites share a common phase boundary, while lattices with an odd number of sites have a distinct phase boundary.

cond-mat.stat-mech↗

arXiv2022

Adma-GAN: Attribute-Driven Memory Augmented GANs for Text-to-Image Generation

As a challenging task, text-to-image generation aims to generate photo-realistic and semantically consistent images according to the given text descriptions. Existing methods mainly extract the text information from only one sentence to represent an image and the text representation effects the quality of the generated image well. However, directly utilizing the limited information in one sentence misses some key attribute descriptions, which are the crucial factors to describe an image accurately. To alleviate the above problem, we propose an effective text representation method with the complements of attribute information. Firstly, we construct an attribute memory to jointly control the text-to-image generation with sentence input. Secondly, we explore two update mechanisms, sample-aware and sample-joint mechanisms, to dynamically optimize a generalized attribute memory. Furthermore, we design an attribute-sentence-joint conditional generator learning scheme to align the feature embeddings among multiple representations, which promotes the cross-modal network training. Experimental results illustrate that the proposed method obtains substantial performance improvements on both the CUB (FID from 14.81 to 8.57) and COCO (FID from 21.42 to 12.39) datasets.

cs.CV↗

arXiv2022

F3A-GAN: Facial Flow for Face Animation with Generative Adversarial Networks

Formulated as a conditional generation problem, face animation aims at synthesizing continuous face images from a single source image driven by a set of conditional face motion. Previous works mainly model the face motion as conditions with 1D or 2D representation (e.g., action units, emotion codes, landmark), which often leads to low-quality results in some complicated scenarios such as continuous generation and largepose transformation. To tackle this problem, the conditions are supposed to meet two requirements, i.e., motion information preserving and geometric continuity. To this end, we propose a novel representation based on a 3D geometric flow, termed facial flow, to represent the natural motion of the human face at any pose. Compared with other previous conditions, the proposed facial flow well controls the continuous changes to the face. After that, in order to utilize the facial flow for face editing, we build a synthesis framework generating continuous images with conditional facial flows. To fully take advantage of the motion information of facial flows, a hierarchical conditional framework is designed to combine the extracted multi-scale appearance features from images and motion features from flows in a hierarchical manner. The framework then decodes multiple fused features back to images progressively. Experimental results demonstrate the effectiveness of our method compared to other state-of-the-art methods.

cs.CV↗

arXiv2022

D3T-GAN: Data-Dependent Domain Transfer GANs for Few-shot Image Generation

As an important and challenging problem, few-shot image generation aims at generating realistic images through training a GAN model given few samples. A typical solution for few-shot generation is to transfer a well-trained GAN model from a data-rich source domain to the data-deficient target domain. In this paper, we propose a novel self-supervised transfer scheme termed D3T-GAN, addressing the cross-domain GANs transfer in few-shot image generation. Specifically, we design two individual strategies to transfer knowledge between generators and discriminators, respectively. To transfer knowledge between generators, we conduct a data-dependent transformation, which projects and reconstructs the target samples into the source generator space. Then, we perform knowledge transfer from transformed samples to generated samples. To transfer knowledge between discriminators, we design a multi-level discriminant knowledge distillation from the source discriminator to the target discriminator on both the real and fake samples. Extensive experiments show that our method improve the quality of generated images and achieves the state-of-the-art FID scores on commonly used datasets.

cs.CV↗

arXiv2022

Quasi time crystal

We discuss the possibility of making a quasi time crystal. A simple two-state model is studied to clarify our definition. In a superposition of the ground state and the excited state and the probability of observation varies periodically in time during the lifetime of the excited state. The quasi time crystal is also discussed around the first order quantum phase transition, which is characterized by the degeneracy and crossing of the two lowest-energy states in the infinite-volume limit. Our results have broad validity. As an example, the one-dimensional transverse field Ising model with surface fields is shown to have similar behavior. The oscillating magnetization profile is solved exactly.

cond-mat.stat-mech↗

arXiv2021

MGH: Metadata Guided Hypergraph Modeling for Unsupervised Person Re-identification

As a challenging task, unsupervised person ReID aims to match the same identity with query images which does not require any labeled information. In general, most existing approaches focus on the visual cues only, leaving potentially valuable auxiliary metadata information (e.g., spatio-temporal context) unexplored. In the real world, such metadata is normally available alongside captured images, and thus plays an important role in separating several hard ReID matches. With this motivation in mind, we propose~\textbf{MGH}, a novel unsupervised person ReID approach that uses meta information to construct a hypergraph for feature learning and label refinement. In principle, the hypergraph is composed of camera-topology-aware hyperedges, which can model the heterogeneous data correlations across cameras. Taking advantage of label propagation on the hypergraph, the proposed approach is able to effectively refine the ReID results, such as correcting the wrong labels or smoothing the noisy labels. Given the refined results, We further present a memory-based listwise loss to directly optimize the average precision in an approximate manner. Extensive experiments on three benchmarks demonstrate the effectiveness of the proposed approach against the state-of-the-art.

cs.CV↗

arXiv2021

Extend the FFmpeg Framework to Analyze Media Content

This paper introduces a new set of video analytics plugins developed for the FFmpeg framework. Multimedia applications that increasingly utilize the FFmpeg media features for its comprehensive media encoding, decoding, muxing, and demuxing capabilities can now additionally analyze the video content based on AI models. The plugins are thread optimized for best performance overcoming certain FFmpeg threading limitations. The plugins utilize the Intel OpenVINO Toolkit inference engine as the backend. The analytics workloads are accelerated on different platforms such as CPU, GPU, FPGA or specialized analytics accelerators. With our reference implementation, the feature of OpenVINO as inference backend has been pushed into FFmpeg mainstream repository. We plan to submit more patches later.

cs.MM↗

arXiv2020

Semantic Neighborhood-Aware Deep Facial Expression Recognition

Different from many other attributes, facial expression can change in a continuous way, and therefore, a slight semantic change of input should also lead to the output fluctuation limited in a small scale. This consistency is important. However, current Facial Expression Recognition (FER) datasets may have the extreme imbalance problem, as well as the lack of data and the excessive amounts of noise, hindering this consistency and leading to a performance decreasing when testing. In this paper, we not only consider the prediction accuracy on sample points, but also take the neighborhood smoothness of them into consideration, focusing on the stability of the output with respect to slight semantic perturbations of the input. A novel method is proposed to formulate semantic perturbation and select unreliable samples during training, reducing the bad effect of them. Experiments show the effectiveness of the proposed method and state-of-the-art results are reported, getting closer to an upper limit than the state-of-the-art methods by a factor of 30\% in AffectNet, the largest in-the-wild FER database by now.

cs.CV↗

arXiv2020

Wetting transition in the McCoy-Wu model

The wetting transition is studied in the McCoy-Wu Ising model in which the random bonds are perfectly correlated in the direction parallel to the walls . The model is solved numerically on finite size lattices up to $200 \times 200^2$. It is shown that the wetting transition is first-order. For a fixed surface field, the distribution of wetting transition temperature is obtained from $1000$ samples. The results show that the deviation of the wetting transition temperature does not decreases as the lattice size increases. It is shown that for a fixed surface field the wetting transition temperature is sample dependent even in the thermodynamic limit.

cond-mat.stat-mech↗

arXiv2016

The Landau-Ginzburg-Wilson Hamiltonian for the Griffiths phase

The Landau-Ginzburg-Wilson Hamiltonian with random temperature for the phase transition in disordered systems from the Griffiths phase to ferromagnetic phase is reexamined. From the saddle point solutions, especially the excited state solutions, it is shown that the system self-organizes into blocks coupled with their neighbors like superspins, which are emergent variables. Taking the fluctuation around these saddle point solutions into account, we get an effective Hamiltonian, including the emergent superspins of the blocks, the fluctuation around the saddle point solutions, and their couplings. Applying Stratonovich-Hubbard transformation to the part of superspins, we get a Landau-Ginzburg-Wilson Hamiltonian for the blocks. From the saddle point equations for the blocks, we can get the second generation blocks, of which sizes are much larger than the first generation blocks. Repeating this procedure again and again, we get many generations of blocks to describe the asymptotic behavior. If a field is applied, the effective field on the superspins is multiplied greatly and proportional to the block size. For a very small field, the effective field on the higher generation superspins can be so strong to cause the superspins polarizaed radically. This can explain the extra large critical isotherm exponent discovered in the experiments. The phase space of reduced temperature vs. field is divided into many layers , in which different generation blocks dominate the critical behavior. The sizes of the different generation emergent blocks are new relevant length scales. This can explain a lot of puzzles in the experiments and the Monte Carlo simulation.

cond-mat.stat-mech↗

arXiv2015

The critical phenomena of a single defect

We consider the critical system with a point defect and study the variation of thermodynamic quantities, which are the differences between those with and without the defect. Within renormalization group theory, we show generally that the critical exponent of the internal energy variation is the specific heat exponent of a pure system, and the critical exponent of the heat capacity variation is that for the temperature derivative of specific heat of a pure system. This conclusion is valid for the isotropic systems with a short-range interaction. As an example we solve the two dimensional Ising model with a point defect numerically. The variations of the free energy, internal energy and specific heat are calculated with bond propagation algorithm. At the critical point, the internal energy variation diverges with the lattice size logarithmically and the heat capacity variation diverges with size linearly. Near the critical point, the internal energy variation behaves as $\ln |t|$ and the heat capacity variation behaves as $|t|^{-1}$, where $t$ is the reduced temperature.

cond-mat.stat-mech↗

arXiv2014

The critical 2-dimensional Ising model with fixed boundaries

The critical 2-dimensional Ising model is studied with four types boundary conditions: free, fixed ferromagnetic, fixed antiferromagnetic, fixed double antiferromagnetic. Using Bond Propagation algorithms with surface fields, we obtained the free energy, internal energy and specific heat numerically on square lattices with square shape and various combinations of the four types boundary conditions. The numerical data are analyzed with finite size scaling. The bulk, edge and corner terms are extracted very accurately. The exact results are conjectured for the corner logarithmic term in the free energy, the edge and corner logarithmic terms in the internal energy and specific heat. The corner logarithmic terms in the free energy agree with the conformal field theory very well.

cond-mat.stat-mech↗

arXiv2014

Efficient algortihms for the two dimensional Ising model with a surface field

Bond propagation and site propagation algorithm are extended to the two dimensional Ising model with a surface field. With these algorithms we can calculate the free energy, internal energy, specific heat, magnetization, correlation function, surface magnetization, surface susceptibility and surface correlation. To test these algorithms, we study the Ising model for wetting transition, which is solved exactly by Abraham. We can locate the transition point accurately to $10^{-8}$. We carry out the calculation of the specific heat, surface susceptibility on the lattices with the sizes are up to $200^2 \times 200$. The results show that finite jump develops in the specific heat and surface susceptibility at the transition point as the lattice size increases. On the lattice with size $320^2 \times 320$ the parallel correlation length exponent is $1.88$, while in the Abraham's exact result it is $2.0$. The perpendicular correlation length exponent on the lattice with size $160^2\times 160$ is $1.04$, where its exact value is $1.0$.

cond-mat.stat-mech↗

arXiv2014

Exact finite-size corrections and corner free energies for the c=-2 universality class

We consider (a) the partition functions of the anisotropic dimer model on the rectangular (2M-1) x (2N-1) lattice with free and cylindrical boundary conditions with a single monomer residing on the boundary and (b) the partition function of the anisotropic spanning tree on an M x N rectangular lattice with free boundary conditions. We express (a) and (b) in terms of a principal partition function with twisted boundary conditions. Based on these expressions, we derive the exact asymptotic expansions of the free energy for both cases (a) and (b). We confirm the conformal field theory prediction for the corner free energy of these models, and find the central charge is c = - 2. We also show that the dimer model on the cylinder with an odd number of sites on the perimeter exhibits the same finite-size corrections as on the plane.

cond-mat.stat-mech↗

arXiv2013

Accurate expansions of internal energy and specific heat of critical two-dimensional Ising model with free boundaries

The bond-propagation (BP) algorithm for the specific heat of the two dimensional Ising model is developed and that for the internal energy is completed. Using these algorithms, we study the critical internal energy and specific heat of the model on the square lattice and triangular lattice with free boundaries. Comparing with previous works [X.-T. Wu {\it et al} Phys. Rev. E {\bf 86}, 041149 (2012) and Phys. Rev. E {\bf 87}, 022124 (2013)], we reach much higher accuracy ($10^{-26}$) of the internal energy and specific heat, compared to the accuracy $10^{-11}$ of the internal energy and $10^{-9}$ of the specific heat reached in the previous works. This leads to much more accurate estimations of the edge and corner terms. The exact values of some edge and corner terms are therefore conjectured. The accurate forms of finite-size scaling for the internal energy and specific heat are determined for the rectangle-shaped square lattice with various aspect ratios and various shaped triangular lattice. For the rectangle-shaped square and triangular lattices and the triangle-shaped triangular lattice, there is no logarithmic correction terms of order higher than 1/S, with S the area of the system. For the triangular lattice in rhombus, trapezoid and hexagonal shapes, there exist logarithmic correction terms of order higher than 1/S for the internal energy, and logarithmic correction terms of all orders for the specific heat.

cond-mat.stat-mech↗

arXiv2012

Shape dependent finite-size effect of critical two-dimensional Ising model on a triangular lattice

Using the bond-propagation algorithm, we study the finite-size behavior of the critical two-dimensional Ising model on a finite triangular lattice with free boundaries in five shapes: triangle, rhombus, trapezoid, hexagon and rectangle. The critical free energy, internal energy and specific heat are calculated. The accuracy of the free energy reaches $10^{-26}$. Based on accurate data on several finite systems with linear size up to N=2000, we extract the bulk, surface and corner parts of the free energy, internal energy and specific heat accurately. We confirm the conformal field theory prediction of the corner free energy to be universal and find logarithmic corrections in higher order terms in the critical free energy for the rhombus, trapezoid, and hexagon shaped systems, which are absent for the triangle and rectangle shaped systems. The logarithmic edge corrections due to edges parallel or perpendicular to the bond directions in the internal energy are found to be identical, while the logarithmic edge corrections due to corresponding edges in the free energy and the specific heat are different. The corner internal energy and corner specific heat for angles $π/3$, $π/2$ and $2π/3$ are obtained, as well as higher order corrections. Comparing with the corner internal energy and corner specific heat previously found on a rectangle of the square lattice (Phys. Rev. E. 86 041149 (2012)), we conclude that the corner internal energy and corner specific heat for the rectangle shape are not universal.

cond-mat.stat-mech↗

arXiv2012

Finite size behaviors of critical Ising model on a rectangle with free boundaries

Using the bond-propagation algorithm, we study the Ising model on a rectangle of size $M \times N$ with free boundaries. For five aspect ratios $ρ=M/N=1,2,4,8,16$, the critical free energy, internal energy and specific heat are calculated. The largest size reached is $M \times N=64\times 10^6$. The accuracy of the free energy reaches $10^{-26}$. Basing on these accurate data, we determine exact expansions of the critical free energy, internal energy and specific heat. With these expansions, we extract the bulk, surface and corner parts of free energy, internal energy and specific heat. The fitted bulk free energy density is given by $f_{\infty}=0.92969539834161021499(1)$, comparing with Onsager's exact result $f_{\infty}=0.92969539834161021506...$. We prove the conformal field theory(CFT) prediction of the corner free energy, in which the central charge of the Ising model is found to be $c=0.5\pm 1\times 10^{-10}$ comparing with the CFT result $c=0.5$. We find that not only the corner free energy but also the corner internal energy and specific heat are geometry independent, i.e., independent of aspect ratio. The implication of this finding on the finite scaling is discussed. In the second order correction of the free energy, we prove the geometry dependence predicted by CFT and find out a geometry independent constant beyond CFT. High order corrections are also obtained.

cond-mat.stat-mech↗