图神经网络是一种强大的学习图数据的模型,通过节点信息嵌入和图卷积运算实现图结构数据的表示。图数据中节点的结构信息和节点的位置信息对获取图特征至关重要,但现有的图神经网络同时捕获位置信息和结构信息的表达能力有限。对此,提...图神经网络是一种强大的学习图数据的模型,通过节点信息嵌入和图卷积运算实现图结构数据的表示。图数据中节点的结构信息和节点的位置信息对获取图特征至关重要,但现有的图神经网络同时捕获位置信息和结构信息的表达能力有限。对此,提出了一种新的图神经网络——融合位置和结构信息的图神经网络(Positional and Structural Information with Graph Neural Networks, PSI-GNN)。PSI-GNN的核心思想在于利用编码器获取节点的位置和结构信息,并将这些信息特征嵌入到网络中。通过在网络中更新和传递这两种信息,PSI-GNN实现了对位置和结构信息的有效融合与利用,为解决上述问题提供了有效的解决方案。同时,为应对不同类型的图学习任务,PSI-GNN给予位置和结构信息以不同的权重来应对不同的下游任务。为了验证PSI-GNN的有效性,在多个基准图数据集上进行了实验。实验结果表明,PSI-GNN在节点级任务上最高提升了约14%,在图级任务上最高提升了约35%,验证了PSI-GNN在同时捕获位置和结构信息方面的有效性。展开更多
Following the fundamental characteristics of the porosity windbreak,this study suggests a new numerical investigation method for the wind field of the windbreak based on the porous medium physical model.This method ca...Following the fundamental characteristics of the porosity windbreak,this study suggests a new numerical investigation method for the wind field of the windbreak based on the porous medium physical model.This method can transform the reasonable matching problem of the porosity and windproof performance of the windbreak into a study of the relationship between the resistance coefficient of the porous medium and the aerodynamic load of the train.This study examines the influence of the hole type on the wind field behind the porosity windbreak.Then,the relationship between the resistance coefficient of the porous medium,the porosity of the windbreak,and the aerodynamic loads of the train is investigated.The results show that the porous media physical model can be used instead of the windbreak geometry to study the windbreak-train aerodynamic performance,and the process of using this method is suggested.展开更多
Beidou-3 navigation satellite system(BDS-3)initiated a real-time service for precise point positioning(PPP)using the B2b signal,mainly for users in China and surrounding areas.In this paper,the performance of PPP-B2b ...Beidou-3 navigation satellite system(BDS-3)initiated a real-time service for precise point positioning(PPP)using the B2b signal,mainly for users in China and surrounding areas.In this paper,the performance of PPP-B2b service is experimentally analyzed first.Then,the ionosphere-free model is established.In order to solve the problem of slow convergence for traditional PPP,an adaptive robust extend Kalman filter(AREKF)algorithm is developed.Unlike the error compensation models,it reflects the noise information in real time by adjusting the covariance matrix of the measurements and the weight matrix of the state vector.The experimental results are analyzed last.Evaluation results indicate that the corrections provided by PPP-B2b can significantly reduce the discontinuous error of the orbits and clock offsets caused by broadcast ephemeris updating.Positioning results confirm that AREKF outperforms EKF both in static and kinematic modes.Around 20%improvement in accuracy and 25%improvement in convergence speed are achieved,making it valuable for PPP processing.展开更多
文摘图神经网络是一种强大的学习图数据的模型,通过节点信息嵌入和图卷积运算实现图结构数据的表示。图数据中节点的结构信息和节点的位置信息对获取图特征至关重要,但现有的图神经网络同时捕获位置信息和结构信息的表达能力有限。对此,提出了一种新的图神经网络——融合位置和结构信息的图神经网络(Positional and Structural Information with Graph Neural Networks, PSI-GNN)。PSI-GNN的核心思想在于利用编码器获取节点的位置和结构信息,并将这些信息特征嵌入到网络中。通过在网络中更新和传递这两种信息,PSI-GNN实现了对位置和结构信息的有效融合与利用,为解决上述问题提供了有效的解决方案。同时,为应对不同类型的图学习任务,PSI-GNN给予位置和结构信息以不同的权重来应对不同的下游任务。为了验证PSI-GNN的有效性,在多个基准图数据集上进行了实验。实验结果表明,PSI-GNN在节点级任务上最高提升了约14%,在图级任务上最高提升了约35%,验证了PSI-GNN在同时捕获位置和结构信息方面的有效性。
基金Projects(52302447,52388102,52372369)supported by the National Natural Science Foundation of China。
文摘Following the fundamental characteristics of the porosity windbreak,this study suggests a new numerical investigation method for the wind field of the windbreak based on the porous medium physical model.This method can transform the reasonable matching problem of the porosity and windproof performance of the windbreak into a study of the relationship between the resistance coefficient of the porous medium and the aerodynamic load of the train.This study examines the influence of the hole type on the wind field behind the porosity windbreak.Then,the relationship between the resistance coefficient of the porous medium,the porosity of the windbreak,and the aerodynamic loads of the train is investigated.The results show that the porous media physical model can be used instead of the windbreak geometry to study the windbreak-train aerodynamic performance,and the process of using this method is suggested.
文摘Beidou-3 navigation satellite system(BDS-3)initiated a real-time service for precise point positioning(PPP)using the B2b signal,mainly for users in China and surrounding areas.In this paper,the performance of PPP-B2b service is experimentally analyzed first.Then,the ionosphere-free model is established.In order to solve the problem of slow convergence for traditional PPP,an adaptive robust extend Kalman filter(AREKF)algorithm is developed.Unlike the error compensation models,it reflects the noise information in real time by adjusting the covariance matrix of the measurements and the weight matrix of the state vector.The experimental results are analyzed last.Evaluation results indicate that the corrections provided by PPP-B2b can significantly reduce the discontinuous error of the orbits and clock offsets caused by broadcast ephemeris updating.Positioning results confirm that AREKF outperforms EKF both in static and kinematic modes.Around 20%improvement in accuracy and 25%improvement in convergence speed are achieved,making it valuable for PPP processing.
