TopoSurfel: Closing the Loop between Gaussian Surfels and Meshes for Surface Reconstruction
为解决3D高斯点云直接提取高质量表面的难题，提出TopoSurfel框架，通过动态生成连续代理网格及网格引导的点云进化策略，有效抑制了伪影和浮动物体。
0 citationsRead paper
Selected work
Representative Papers
The Fourier Spectral Transformer Networks For Efficient and Generalizable Nonlinear PDEs Prediction
To address the low accuracy and poor generalization of long-term predictions for nonlinear partial differential equations (PDEs) under limited training data, this paper proposes a spectral-domain Transformer framework. First, the PDE is transformed via Fourier spectral decomposition into a system of ordinary differential equations (ODEs) governing the temporal evolution of spectral coefficients. High-order ODE solvers are then employed to generate high-fidelity synthetic training data. Finally, a lightweight Transformer architecture is designed to model the spatiotemporal dynamics of these spectral coefficients. The method innovatively integrates the physical fidelity of classical spectral methods with the long-range dependency modeling capability of self-attention mechanisms. Experiments on the two-dimensional incompressible Navier–Stokes equations and the one-dimensional Burgers equation demonstrate superior long-term prediction accuracy and generalization over traditional numerical schemes and state-of-the-art data-driven models—particularly under severe data scarcity.
0 citationsRead paper
The geomagnetic storm and Kp prediction using Wasserstein transformer
This study addresses the need for accurate geomagnetic storm forecasting by proposing a multimodal Wasserstein Transformer framework for high-precision 3-day and 5-day planetary Kp index prediction. Methodologically, it introduces Wasserstein distance into both the Transformer encoder architecture and the loss function—marking the first such integration—to jointly align cross-modal probability distributions and model temporal dynamics. The framework fuses heterogeneous data sources: satellite measurements, solar extreme ultraviolet (EUV) imagery, and historical Kp time series. Compared to the NOAA operational model, it achieves superior consistency during quiet periods and enhanced fidelity across all storm phases, demonstrating greater robustness and practicality in real-time forecasting. The core contribution lies in a Wasserstein-driven probabilistic alignment mechanism for multisource, heterogeneous data—establishing an interpretable and generalizable deep learning paradigm for long-term geomagnetic activity prediction.
0 citationsRead paper
Recent publications
Latest Papers
TopoSurfel: Closing the Loop between Gaussian Surfels and Meshes for Surface Reconstruction
为解决3D高斯点云直接提取高质量表面的难题，提出TopoSurfel框架，通过动态生成连续代理网格及网格引导的点云进化策略，有效抑制了伪影和浮动物体。
0 citationsRead paper
The Fourier Spectral Transformer Networks For Efficient and Generalizable Nonlinear PDEs Prediction
To address the low accuracy and poor generalization of long-term predictions for nonlinear partial differential equations (PDEs) under limited training data, this paper proposes a spectral-domain Transformer framework. First, the PDE is transformed via Fourier spectral decomposition into a system of ordinary differential equations (ODEs) governing the temporal evolution of spectral coefficients. High-order ODE solvers are then employed to generate high-fidelity synthetic training data. Finally, a lightweight Transformer architecture is designed to model the spatiotemporal dynamics of these spectral coefficients. The method innovatively integrates the physical fidelity of classical spectral methods with the long-range dependency modeling capability of self-attention mechanisms. Experiments on the two-dimensional incompressible Navier–Stokes equations and the one-dimensional Burgers equation demonstrate superior long-term prediction accuracy and generalization over traditional numerical schemes and state-of-the-art data-driven models—particularly under severe data scarcity.
0 citationsRead paper
The geomagnetic storm and Kp prediction using Wasserstein transformer
This study addresses the need for accurate geomagnetic storm forecasting by proposing a multimodal Wasserstein Transformer framework for high-precision 3-day and 5-day planetary Kp index prediction. Methodologically, it introduces Wasserstein distance into both the Transformer encoder architecture and the loss function—marking the first such integration—to jointly align cross-modal probability distributions and model temporal dynamics. The framework fuses heterogeneous data sources: satellite measurements, solar extreme ultraviolet (EUV) imagery, and historical Kp time series. Compared to the NOAA operational model, it achieves superior consistency during quiet periods and enhanced fidelity across all storm phases, demonstrating greater robustness and practicality in real-time forecasting. The core contribution lies in a Wasserstein-driven probabilistic alignment mechanism for multisource, heterogeneous data—establishing an interpretable and generalizable deep learning paradigm for long-term geomagnetic activity prediction.
0 citationsRead paper