arXiv subjects

Hongyu Lu

Publications and source records attributed to Hongyu Lu.

43 records · Page 3Linked to original sources

arXiv2022

Thermodynamic characteristic for correlated flat-band system with quantum anomalous Hall ground state

While the ground state phase diagram of the correlated flat-band systems have been intensively investigated, the dynamic and thermodynamic properties of such lattice models are less explored, but it is the latter which is most relevant to the experimental probes (transport, quantum capacitance and spectroscopy) of the quantum moiré materials such as twisted bilayer graphene and transition metal dichalcogenides. Here we show, by means of momentum-space quantum Monte Carlo and exact diagonalization, there exists a unique thermodynamic characteristic for the correlated flat-band models with interaction-driven quantum anomalous Hall (QAH) ground state, namely, the transition from the QAH insulator to the metallic state takes place at a much lower temperature compared with the zero-temperature single-particle gap generated by the long-range Coulomb interaction. Such low transition temperature comes from the proliferation of excitonic particle-hole excitations, which "quantum teleport" the electrons across the gap between different topological bands to restore the broken time-reversal symmetry and give rise to a pronounced enhancement in the charge compressibility. Future experiments, to verify such generic thermodynamic characteristics, are proposed.

cond-mat.str-el↗

arXiv2022

Network-Initialized Monte Carlo Based on Generative Neural Networks

We design generative neural networks that generate Monte Carlo configurations with complete absence of autocorrelation from which only short Markov chains are needed before making measurements for physical observables, irrespective of the system locating at the classical critical point, fermionic Mott insulator, Dirac semimetal, or quantum critical point. We further propose a network-initialized Monte Carlo scheme based on such neural networks, which provides independent samplings and can accelerate the Monte Carlo simulations by significantly reducing the thermalization process. We demonstrate the performance of our approach on the two-dimensional Ising and fermion Hubbard models, and expect it can systematically speed up the Monte Carlo simulations especially for the very challenging many-electron problems.

cond-mat.str-el↗

arXiv2019

Improving Visual Recognition using Ambient Sound for Supervision

Our brains combine vision and hearing to create a more elaborate interpretation of the world. When the visual input is insufficient, a rich panoply of sounds can be used to describe our surroundings. Since more than 1,000 hours of videos are uploaded to the internet everyday, it is arduous, if not impossible, to manually annotate these videos. Therefore, incorporating audio along with visual data without annotations is crucial for leveraging this explosion of data for recognizing and understanding objects and scenes. Owens,et.al suggest that a rich representation of the physical world can be learned by using a convolutional neural network to predict sound textures associated with a given video frame. We attempt to reproduce the claims from their experiments, of which the code is not publicly available. In addition, we propose improvements in the pretext task that result in better performance in other downstream computer vision tasks.

cs.CV↗

arXiv2016

In Situ Simultaneous Measuring Forces Acted on Individual Legs of Water Striders According to Archimedes' Principle

The supporting forces of all the legs have been simultaneously measured through a natural phenomenon inspired shadow method, by capturing water strider leg shadows to calculate the expelled water volume and the equivalent floating forces according to Archimedes' principle, with a sensitivity up to nN~pN/pixel. The method was verified by comparing to electronic balance to weight water striders. The supporting force acted on legs was found to linearly proportional to its shadow area, which geometry clearly indicated the hydrophobicity of legs. The pressed depth of legs, vertical weight focus position change and body pitch angle during leg refreshing, sculling and jumping forward of water striders have been achieved with resolutions of 5um/pixel, 2 um/pixel, and 0.02 degree, respectively, which is difficult for general imaging technologies. The shadow method was also developed into a force measuring device and characterized the adhesion force of a hair/polymer surface contact in a few uN with a 0.3 nN/pixel resolution.

cond-mat.soft↗

arXiv2012

A United Image Force for Deformable Models and Direct Transforming Geometric Active Contorus to Snakes by Level Sets

A uniform distribution of the image force field around the object fasts the convergence speed of the segmentation process. However, to achieve this aim, it causes the force constructed from the heat diffusion model unable to indicate the object boundaries accurately. The image force based on electrostatic field model can perform an exact shape recovery. First, this study introduces a fusion scheme of these two image forces, which is capable of extracting the object boundary with high precision and fast speed. Until now, there is no satisfied analysis about the relationship between Snakes and Geometric Active Contours (GAC). The second contribution of this study addresses that the GAC model can be deduced directly from Snakes model. It proves that each term in GAC and Snakes is correspondent and has similar function. However, the two models are expressed using different mathematics. Further, since losing the ability of rotating the contour, adoption of level sets can limits the usage of GAC in some circumstances.

cs.CV↗

arXiv2009

An Improved Algorithm based on Shannon-Happ Formula for Calculating Transfer Function from Signal Flow Graph and Its Visualization

A new method based on Shannon-Happ formula to calculate transfer function from Signal Flow Graph (SFG) is presented. The algorithm provides an explicit approach to get the transfer function in a format with both numerical and symbolic expressions. The adoption of the symbolic variable in SFG, which could represent the nonlinear item or the independent sub-system, is achieved by variable separation approach. An investigation is given for the solutions of several special conditions of SFG. To improve the efficiency of the algorithm, a new technique combined with Johnson method for generating the combinations of the non-touching loops is developed. It uses the previous combinations in lower order to get the ones in higher order. There is an introduction about the visualization of SFG and the subroutines for system performance analysis in the software, AVANT.

math.NA↗

arXiv2009

Physical Modeling Techniques in Active Contours for Image Segmentation

Physical modeling method, represented by simulation and visualization of the principles in physics, is introduced in the shape extraction of the active contours. The objectives of adopting this concept are to address the several major difficulties in the application of Active Contours. Primarily, a technique is developed to realize the topological changes of Parametric Active Contours (Snakes). The key strategy is to imitate the process of a balloon expanding and filling in a closed space with several objects. After removing the touched balloon surfaces, the objects can be identified by surrounded remaining balloon surfaces. A burned region swept by Snakes is utilized to trace the contour and to give a criterion for stopping the movement of Snake curve. When the Snakes terminates evolution totally, through ignoring this criterion, it can form a connected area by evolving the Snakes again and continuing the region burning. The contours extracted from the boundaries of the burned area can represent the child snake of each object respectively. Secondly, a novel scheme is designed to solve the problems of leakage of the contour from the large gaps, and the segmentation error in Geometric Active Contours (GAC). It divides the segmentation procedure into two processing stages. By simulating the wave propagating in the isotropic substance at the final stage, it can significantly enhance the effect of image force in GAC based on Level Set and give the satisfied solutions to the two problems. Thirdly, to support the physical models for active contours above, we introduce a general image force field created on a template plane over the image plane. This force is more adaptable to noisy images with complicated geometric shapes.

cs.CV↗