BZPEERPapers, people, evidence and research paths
Menu

Discover

SearchFind papers, people, topics, institutions and opportunities from one place.GraphSee how papers, authors, topics and evidence connect.

Research tools

HomeContinue from the papers, people and topics that matter now.CollectionsTurn saved papers into reading lists, evidence maps and shareable context.ResearchReconstruct the path that led you from search to paper to question.CollaborationMove from a paper, topic or expert request into a focused thread.PublishCreate full-text research articles with metadata, formulas and versions.

Network

ExpertsFind specialists whose work and reviews explain why they can help.CommunitiesJoin research groups organized around papers, topics and shared evidence.

Opportunities

ListingsBrowse roles, calls and collaborations connected to real research context.

Account

Log inReturn to your saved papers, graph views and active threads.Create accountStart saving papers, building a research trail and joining teams.
Log inCreate account

Source author record

Shan Zhao

Shan Zhao appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record
math.NAComputer Visioncond-mat.mtrl-sciphysics.comp-phArtificial IntelligenceBiomoleculesComputation and Languagecond-mat.mes-hallcond-mat.supr-conMachine LearningMolecular Networks

Catalog footprint

What is connected

14works
11topics
4close collaborators

Actions

Connect this record

Log in to claim
Open graphBrowse works

Research graph

See the researcher in context

Open full explorer

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

14 published item(s)

preprint2026arXiv

ParkGaussian: Surround-view 3D Gaussian Splatting for Autonomous Parking

Parking is a critical task for autonomous driving systems (ADS), with unique challenges in crowded parking slots and GPS-denied environments. However, existing works focus on 2D parking slot perception, mapping, and localization, 3D reconstruction remains underexplored, which is crucial for capturing complex spatial geometry in parking scenarios. Naively improving the visual quality of reconstructed parking scenes does not directly benefit autonomous parking, as the key entry point for parking is the slots perception module. To address these limitations, we curate the first benchmark named ParkRecon3D, specifically designed for parking scene reconstruction. It includes sensor data from four surround-view fisheye cameras with calibrated extrinsics and dense parking slot annotations. We then propose ParkGaussian, the first framework that integrates 3D Gaussian Splatting (3DGS) for parking scene reconstruction. To further improve the alignment between reconstruction and downstream parking slot detection, we introduce a slot-aware reconstruction strategy that leverages existing parking perception methods to enhance the synthesis quality of slot regions. Experiments on ParkRecon3D demonstrate that ParkGaussian achieves state-of-the-art reconstruction quality and better preserves perception consistency for downstream tasks. The code and dataset will be released at: https://github.com/wm-research/ParkGaussian

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Xiaomi EV World Model: A Joint World Model Integrating Reconstruction and Generation for Autonomous Driving

This report presents a unified technical system addressing the two core capabilities of world models for autonomous driving: world representation and world generation. For world representation, we propose WorldRec, a feed-forward reconstruction architecture driven by sparse scene queries. WorldRec initializes structured queries in 3D space, leveraging them to aggregate cross-view, cross-temporal features, thereby naturally enforcing spatial consistency across frames and yielding compact yet high-fidelity 3D Gaussian scene representations. For world generation, we propose WorldGen, a two-stage training framework of bidirectional pretraining followed by causal fine-tuning through three progressive stages (Teacher Forcing, ODE distillation, and DMD), enabling high-quality online causal video generation in as few as 4 denoising steps. Building on both modules, we further introduce the JWM, which deeply integrates WorldRec and WorldGen to achieve synergistic gains in generation stability, cross-frame consistency, and visual fidelity, providing a solid foundation for closed-loop simulation, data synthesis, and end-to-end training in autonomous driving.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Reiterative Domain Aware Multi-Target Adaptation

Most domain adaptation methods focus on single-source-single-target adaptation settings. Multi-target domain adaptation is a powerful extension in which a single classifier is learned for multiple unlabeled target domains. To build a multi-target classifier, it is important to have: a feature extractor that generalizes well across domains; and effective aggregation of features from the labeled source and different unlabeled target domains. Towards the first, we use the recently popular Transformer as a feature extraction backbone. Towards the second, we use a co-teaching-based approach using a dual-classifier head, one of which is based on the graph neural network. The proposed approach uses a sequential adaptation strategy that adapts one domain at a time starting from the target domains that are more similar to the source, assuming that the network finds it easier to adapt to such target domains. After adapting on each target, samples with a softmax-based confidence score greater than a threshold are added to the pseudo-source, thus aggregating knowledge from different domains. However, softmax is not entirely trustworthy as a confidence score and may generate a high score for unreliable samples if trained for many iterations. To mitigate this effect, we adopt a reiterative approach, where we reduce target adaptation iterations, however, reiterate multiple times over the target domains. The experimental evaluation on the Office-Home, Office-31 and DomainNet datasets shows significant improvement over the existing methods. We have achieved 10.7$\%$ average improvement in Office-Home dataset over the state-of-art methods.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Revealing sign-reversal $s^{+-}$-wave pairing by quasiparticle interference in the heavy-fermion superconductor CeCu$_2$Si$_2$

Recent observations of two nodeless gaps in superconducting CeCu$_2$Si$_2$ have raised intensive debates as to its exact gap structure of either sign-reversal ($s^{+-}$) or sign-preserving ($s^{++}$) pairing. Here we investigate the quasiparticle interference (QPI) using realistic Fermi surface topology for both weak and strong interband impurity scatterings. Our calculations of the QPI and integrated antisymmetrized local density of states reveal qualitative distinctions between $s^{+-}$ and $s^{++}$ pairing states, which include the intragap impurity resonance and a significant energy-dependence difference between two gap energies. Our predictions provide a guide for phase-sensitive QPI measurements to uncover decisively the true pairing symmetry in the heavy-fermion superconductor CeCu$_2$Si$_2$.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Semi-Supervised Wide-Angle Portraits Correction by Multi-Scale Transformer

We propose a semi-supervised network for wide-angle portraits correction. Wide-angle images often suffer from skew and distortion affected by perspective distortion, especially noticeable at the face regions. Previous deep learning based approaches need the ground-truth correction flow maps for training guidance. However, such labels are expensive, which can only be obtained manually. In this work, we design a semi-supervised scheme and build a high-quality unlabeled dataset with rich scenarios, allowing us to simultaneously use labeled and unlabeled data to improve performance. Specifically, our semi-supervised scheme takes advantage of the consistency mechanism, with several novel components such as direction and range consistency (DRC) and regression consistency (RC). Furthermore, different from the existing methods, we propose the Multi-Scale Swin-Unet (MS-Unet) based on the multi-scale swin transformer block (MSTB), which can simultaneously learn short-distance and long-distance information to avoid artifacts. Extensive experiments demonstrate that the proposed method is superior to the state-of-the-art methods and other representative baselines. The source code and dataset are available at: https://github.com/megvii-research/Portraits_Correction.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

The Utility of General Domain Transfer Learning for Medical Language Tasks

The purpose of this study is to analyze the efficacy of transfer learning techniques and transformer-based models as applied to medical natural language processing (NLP) tasks, specifically radiological text classification. We used 1,977 labeled head CT reports, from a corpus of 96,303 total reports, to evaluate the efficacy of pretraining using general domain corpora and a combined general and medical domain corpus with a bidirectional representations from transformers (BERT) model for the purpose of radiological text classification. Model performance was benchmarked to a logistic regression using bag-of-words vectorization and a long short-term memory (LSTM) multi-label multi-class classification model, and compared to the published literature in medical text classification. The BERT models using either set of pretrained checkpoints outperformed the logistic regression model, achieving sample-weighted average F1-scores of 0.87 and 0.87 for the general domain model and the combined general and biomedical-domain model. General text transfer learning may be a viable technique to generate state-of-the-art results within medical NLP tasks on radiological corpora, outperforming other deep models such as LSTMs. The efficacy of pretraining and transformer-based models could serve to facilitate the creation of groundbreaking NLP models in the uniquely challenging data environment of medical text.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Modulation of heat transport in two-dimensional group-III chalcogenides

We systematically investigated the modulation of heat transport of experimentally accessible two-dimensional (2D) group-III chalcogenides by firstprinciples calculations. It was found that intrinsic thermal conductivity (kappa) of chalcogenides MX (M = Ga, In; X = S, Se) were desirable for efficient heat dissipation. Meanwhile, we showed that the long-range anharmonic interactions played an important role in heat transport of the chalcogenides. The difference of kappa among the 2D group-III chalcogenides can be well described by the Slack model and can be mainly attributed to phonon group velocity. Based on that, we proposed three methods including strain engineering, size effect and making Janus structures to effectively modulate the kappa of 2D group-III chalcogenides, with different underlying mechanisms. We found that tensile strain and rough boundary scattering could continuously decrease the kappa while compressive strain could increase the kappa of 2D group-III chalcogenides. On the other side, the change of kappa by producing Janus structures is permanent and dependent on the structural details. These results provide guilds to modulate heat transport properties of 2D group-III chalcogenides for devices application

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Theoretical study of structure and magnetism of Ga$_{1-x}$V$_x$Sb compounds for spintronic applications

In this paper, the structural, electronic and magnetic properties of Zinc-blende Ga1-xVxSb compounds, with x from dilute doping situation to extreme doping limiting, were systematically investigated by first-principles calculations. V atoms prefer to substitute the Ga atoms and the formation energy is lower in Sb-rich than Ga-rich growth condition. Meantime, the SbGa antisite defects can effectively decrease the energy barrier of substitution process, from 0.85 eV to 0.53 eV. The diffusion of V atom in GaSb lattice is through meta-stable interstitial sites with an energy barrier of 0.6 eV. At a low V concentration x = 0.0625, V atoms prefer a homogeneous distribution and an antiferromagnetic coupling among them. However, starting from x = 0.5, the magnetic coupling among V atoms changes to be ferromagnetic, due to enhanced superexchange interaction between eg and t2g states of neighbouring V atoms. At the extreme limiting of x = 1.00, we found that Zinc-blende VSb as well as its analogs VAs and VP are intrinsic ferromagneitc semiconductors, with a large change of light absorption at the curie temperature. These results indicate that Ga1-xVxSb compounds can provide a platform to design the new electronic, spintronic and optoelectronic devices.

0 citations0 reviewsNo code linkedOpen access
preprint2014arXiv

A matched alternating direction implicit (ADI) method for solving the heat equation with interfaces

A novel Douglas alternating direction implicit (ADI) method is proposed in this work to solve a two-dimensional (2D) heat equation with interfaces. The ADI scheme is a powerful finite difference method for solving parabolic equations, due to its unconditional stability and high efficiency. However, it suffers from a serious accuracy reduction in space for interface problems with different materials and nonsmooth solutions. If the jumps in a function and its derivatives are known across the interface, rigorous ADI schemes have been successfully constructed in the literature based on the immersed interface method (IIM) so that the spatial accuracy can be restored. Nevertheless, the development of accurate and stable ADI methods for general parabolic interface problems with physical interface conditions that describe jumps of a function and its flux, remains unsolved. To overcome this difficulty, a novel tensor product decomposition is proposed in this paper to decouple 2D jump conditions into essentially one-dimensional (1D) ones. These 1D conditions can then be incorporated into the ADI central difference discretization, using the matched interface and boundary (MIB) technique. Fast algebraic solvers for perturbed tridiagonal systems are developed to maintain the computational efficiency. Stability analysis is conducted through eigenvalue spectrum analysis, which numerically demonstrates the unconditional stability of the proposed ADI method. The matched ADI scheme achieves the first order of accuracy in time and second order of accuracy in space in all tested parabolic interface problems with complex geometries and spatial-temporal dependent jump conditions.

0 citations0 reviewsNo code linkedOpen access
preprint2014arXiv

Unconditionally stable time splitting methods for the electrostatic analysis of solvated biomolecules

This work introduces novel unconditionally stable operator splitting methods for solving the time dependent nonlinear Poisson-Boltzmann (NPB) equation for the electrostatic analysis of solvated biomolecules. In a pseudo-transient continuation solution of the NPB equation, a long time integration is needed to reach the steady state. This calls for time stepping schemes that are stable and accurate for large time increments. The existing alternating direction implicit (ADI) methods for the NPB equation are known to be conditionally stable, although being fully implicit. To overcome this difficulty, we propose several new operator splitting schemes, in both multiplicative and additive styles, including locally one-dimensional (LOD) schemes and additive operator splitting (AOS) schemes. The proposed schemes become much more stable than the ADI methods, and some of them are indeed unconditionally stable in dealing with solvated proteins with source singularities and non-smooth solutions. Numerically, the orders of convergence in both space and time are found to be one. Nevertheless, the precision in calculating the electrostatic free energy is low, unless a small time increment is used. Further accuracy improvements are thus considered. After acceleration, the optimized LOD method can produce a reliable energy estimate by integrating for a small and fixed number of time steps. Since one only needs to solve a tridiagonal linear system in each independent one dimensional process, the overall computation is very efficient. The unconditionally stable LOD method scales linearly with respect to the number of atoms in the protein studies, and is over 20 times faster than the conditionally stable ADI methods.

0 citations0 reviewsNo code linkedOpen access
preprint2013arXiv

A Numerical Study on the Weak Galerkin Method for the Helmholtz Equation

A weak Galerkin (WG) method is introduced and numerically tested for the Helmholtz equation. This method is flexible by using discontinuous piecewise polynomials and retains the mass conservation property. At the same time, the WG finite element formulation is symmetric and parameter free. Several test scenarios are designed for a numerical investigation on the accuracy, convergence, and robustness of the WG method in both inhomogeneous and homogeneous media over convex and non-convex domains. Challenging problems with high wave numbers are also examined. Our numerical experiments indicate that the weak Galerkin is a finite element technique that is easy to implement, and provides very accurate and robust numerical solutions for the Helmholtz problem with high wave numbers.

0 citations0 reviewsNo code linkedOpen access
preprint2012arXiv

Weak Galerkin Methods for Second Order Elliptic Interface Problems

Weak Galerkin methods refer to general finite element methods for PDEs in which differential operators are approximated by their weak forms as distributions. Such weak forms give rise to desirable flexibilities in enforcing boundary and interface conditions. A weak Galerkin finite element method (WG-FEM) is developed in this paper for solving elliptic partial differential equations (PDEs) with discontinuous coefficients and interfaces. The paper also presents many numerical tests for validating the WG-FEM for solving second order elliptic interface problems. For such interface problems, the solution possesses a certain singularity due to the nonsmoothness of the interface. A challenge in research is to design high order numerical methods that work well for problems with low regularity in the solution. The best known numerical scheme in the literature is of order one for the solution itself in $L_\infty$ norm. It is demonstrated that the WG-FEM of lowest order is capable of delivering numerical approximations that are of order 1.75 in the usual $L_\infty$ norm for $C^1$ or Lipschitz continuous interfaces associated with a $C^1$ or $H^2$ continuous solutions. Theoretically, it is proved that high order of numerical schemes can be designed by using the WG-FEM with polynomials of high order on each element.

0 citations0 reviewsNo code linkedOpen access
preprint2011arXiv

A Numerical Study on the Weak Galerkin Method for the Helmholtz Equation with Large Wave Numbers

Weak Galerkin (WG) refers to general finite element methods for partial differential equations in which differential operators are approximated by weak forms through the usual integration by parts. In particular, WG methods allow the use of discontinuous finite element functions in the algorithm design. One of such examples was recently introduced by Wang and Ye for solving second order elliptic problems. The goal of this paper is to apply the WG method of Wang and Ye to the Helmholtz equation with high wave numbers. Several test scenarios are designed for a numerical investigation on the accuracy, convergence, and robustness of the WG method in both inhomogeneous and homogeneous media over convex and non-convex domains. Our numerical experiments indicate that weak Galerkin is a finite element technique that is easy to implement, and provides very accurate and robust numerical solutions for the Helmholtz problem with high wave numbers.

0 citations0 reviewsNo code linkedOpen access
preprint2010arXiv

Inferring the Sign of Kinase-Substrate Interactions by Combining Quantitative Phosphoproteomics with a Literature-Based Mammalian Kinome Network

Protein phosphorylation is a reversible post-translational modification commonly used by cell signaling networks to transmit information about the extracellular environment into intracellular organelles for the regulation of the activity and sorting of proteins within the cell. For this study we reconstructed a literature-based mammalian kinase-substrate network from several online resources. The interactions within this directed graph network connect kinases to their substrates, through specific phosphosites including kinase-kinase regulatory interactions. However, the "signs" of links, activation or inhibition of the substrate upon phosphorylation, within this network are mostly unknown. Here we show how we can infer the "signs" indirectly using data from quantitative phosphoproteomics experiments applied to mammalian cells combined with the literature-based kinase-substrate network. Our inference method was able to predict the sign for 321 links and 153 phosphosites on 120 kinases, resulting in signed and directed subnetwork of mammalian kinase-kinase interactions. Such an approach can rapidly advance the reconstruction of cell signaling pathways and networks regulating mammalian cells.

0 citations0 reviewsNo code linkedOpen access