fanxiaohong
My research interest is the application of deep learning in image processing.
Zhejiang Normal University
Pinned Repositories
Breast-Cancer-Dataset
Leveraging Multimodal MRI-based Radiomics Analysis with Diverse Machine Learning Models to Evaluate Lymphovascular Invasion in Clinically Node-Negative Breast Cancer
CGPD-CSNet
Correlation-between-radiation-induced-lung-injury-and-3D-dose-distribution
《放射性肺损伤与三维剂量分布的相关性研究》一文的相关信息
DGDN
MVMS-RCN
Nest-DGIL
Physics-informed-Parallel-Neural-Network-for-Solving-Anisotropic-Elliptic-Interface-Problems
Abstract: In this study, we propose a physics-informed parallel neural network for solving anisotropic elliptic interface problems, which can obtain accurate solutions both near and at the interface. We rewrite the original second-order system into a first-order system to reduce the regularity requirement of solutions by an auxiliary variable. We use different sub-networks to predict solutions in different subdomains and couple them together by loss function terms designed by the interface jump conditions. In addition, we use a piecewise loss function combining L1 norm and L2 norm to enhance the robustness of outliers. Numerical experiment verifies the accuracy and effectiveness of proposed method for solving anisotropic elliptic interface problems. Key Words: anisotropic, elliptic interface, neural network, first-order system
PRISTA-Net
PRM-Net
fanxiaohong's Repositories
fanxiaohong/DGDN
fanxiaohong/Nest-DGIL
fanxiaohong/CGPD-CSNet
fanxiaohong/MVMS-RCN
fanxiaohong/Breast-Cancer-Dataset
Leveraging Multimodal MRI-based Radiomics Analysis with Diverse Machine Learning Models to Evaluate Lymphovascular Invasion in Clinically Node-Negative Breast Cancer
fanxiaohong/Correlation-between-radiation-induced-lung-injury-and-3D-dose-distribution
《放射性肺损伤与三维剂量分布的相关性研究》一文的相关信息
fanxiaohong/Physics-informed-Parallel-Neural-Network-for-Solving-Anisotropic-Elliptic-Interface-Problems
Abstract: In this study, we propose a physics-informed parallel neural network for solving anisotropic elliptic interface problems, which can obtain accurate solutions both near and at the interface. We rewrite the original second-order system into a first-order system to reduce the regularity requirement of solutions by an auxiliary variable. We use different sub-networks to predict solutions in different subdomains and couple them together by loss function terms designed by the interface jump conditions. In addition, we use a piecewise loss function combining L1 norm and L2 norm to enhance the robustness of outliers. Numerical experiment verifies the accuracy and effectiveness of proposed method for solving anisotropic elliptic interface problems. Key Words: anisotropic, elliptic interface, neural network, first-order system
fanxiaohong/PRISTA-Net
fanxiaohong/PRM-Net