The adjoint method is widely used in gradient-based optimization with high-dimensional design variables.However,the cost of solving the adjoint equations in each iteration is comparable to that of solving the flow fie...The adjoint method is widely used in gradient-based optimization with high-dimensional design variables.However,the cost of solving the adjoint equations in each iteration is comparable to that of solving the flow field,resulting in expensive computational costs.To improve the efficiency of solving adjoint equations,we propose a physics-constrained graph neural networks for solving adjoint equations,named ADJ-PCGN.ADJ-PCGN establishes a mapping relationship between flow characteristics and adjoint vector based on data,serving as a replacement for the computationally expensive numerical solution of adjoint equations.A physics-based graph structure and message-passing mechanism are designed to endow its strong fitting and generalization capabilities.Taking transonic drag reduction and maximum lift-drag ratio of the airfoil as examples,results indicate that ADJ-PCGN attains a similar optimal shape as the classical direct adjoint loop method.In addition,ADJ-PCGN demonstrates strong generalization capabilities across different mesh topologies,mesh densities,and out-of-distribution conditions.It holds the potential to become a universal model for aerodynamic shape optimization involving states,geometries,and meshes.展开更多
The increasing popularity of the Internet and the widespread use of information technology have led to a rise in the number and sophistication of network attacks and security threats.Intrusion detection systems are cr...The increasing popularity of the Internet and the widespread use of information technology have led to a rise in the number and sophistication of network attacks and security threats.Intrusion detection systems are crucial to network security,playing a pivotal role in safeguarding networks from potential threats.However,in the context of an evolving landscape of sophisticated and elusive attacks,existing intrusion detection methodologies often overlook critical aspects such as changes in network topology over time and interactions between hosts.To address these issues,this paper proposes a real-time network intrusion detection method based on graph neural networks.The proposedmethod leverages the advantages of graph neural networks and employs a straightforward graph construction method to represent network traffic as dynamic graph-structured data.Additionally,a graph convolution operation with a multi-head attention mechanism is utilized to enhance the model’s ability to capture the intricate relationships within the graph structure comprehensively.Furthermore,it uses an integrated graph neural network to address dynamic graphs’structural and topological changes at different time points and the challenges of edge embedding in intrusion detection data.The edge classification problem is effectively transformed into node classification by employing a line graph data representation,which facilitates fine-grained intrusion detection tasks on dynamic graph node feature representations.The efficacy of the proposed method is evaluated using two commonly used intrusion detection datasets,UNSW-NB15 and NF-ToN-IoT-v2,and results are compared with previous studies in this field.The experimental results demonstrate that our proposed method achieves 99.3%and 99.96%accuracy on the two datasets,respectively,and outperforms the benchmark model in several evaluation metrics.展开更多
Federated Graph Neural Networks (FedGNNs) have achieved significant success in representation learning for graph data, enabling collaborative training among multiple parties without sharing their raw graph data and so...Federated Graph Neural Networks (FedGNNs) have achieved significant success in representation learning for graph data, enabling collaborative training among multiple parties without sharing their raw graph data and solving the data isolation problem faced by centralized GNNs in data-sensitive scenarios. Despite the plethora of prior work on inference attacks against centralized GNNs, the vulnerability of FedGNNs to inference attacks has not yet been widely explored. It is still unclear whether the privacy leakage risks of centralized GNNs will also be introduced in FedGNNs. To bridge this gap, we present PIAFGNN, the first property inference attack (PIA) against FedGNNs. Compared with prior works on centralized GNNs, in PIAFGNN, the attacker can only obtain the global embedding gradient distributed by the central server. The attacker converts the task of stealing the target user’s local embeddings into a regression problem, using a regression model to generate the target graph node embeddings. By training shadow models and property classifiers, the attacker can infer the basic property information within the target graph that is of interest. Experiments on three benchmark graph datasets demonstrate that PIAFGNN achieves attack accuracy of over 70% in most cases, even approaching the attack accuracy of inference attacks against centralized GNNs in some instances, which is much higher than the attack accuracy of the random guessing method. Furthermore, we observe that common defense mechanisms cannot mitigate our attack without affecting the model’s performance on mainly classification tasks.展开更多
Graph Neural Networks(GNNs)have demonstrated outstanding capabilities in processing graph-structured data and are increasingly being integrated into large-scale pre-trained models,such as Large Language Models(LLMs),t...Graph Neural Networks(GNNs)have demonstrated outstanding capabilities in processing graph-structured data and are increasingly being integrated into large-scale pre-trained models,such as Large Language Models(LLMs),to enhance structural reasoning,knowledge retrieval,and memory management.The expansion of their application scope imposes higher requirements on the robustness of GNNs.However,as GNNs are applied to more dynamic and heterogeneous environments,they become increasingly vulnerable to real-world perturbations.In particular,graph data frequently encounters joint adversarial perturbations that simultaneously affect both structures and features,which are significantly more challenging than isolated attacks.These disruptions,caused by incomplete data,malicious attacks,or inherent noise,pose substantial threats to the stable and reliable performance of traditional GNN models.To address this issue,this study proposes the Dual-Shield Graph Neural Network(DSGNN),a defense model that simultaneously mitigates structural and feature perturbations.DSGNN utilizes two parallel GNN channels to independently process structural noise and feature noise,and introduces an adaptive fusion mechanism that integrates information from both pathways to generate robust node representations.Theoretical analysis demonstrates that DSGNN achieves a tighter robustness boundary under joint perturbations compared to conventional single-channel methods.Experimental evaluations across Cora,CiteSeer,and Industry datasets show that DSGNN achieves the highest average classification accuracy under various adversarial settings,reaching 81.24%,71.94%,and 81.66%,respectively,outperforming GNNGuard,GCN-Jaccard,GCN-SVD,RGCN,and NoisyGNN.These results underscore the importance of multi-view perturbation decoupling in constructing resilient GNN models for real-world applications.展开更多
Skin diseases affect millions worldwide.Early detection is key to preventing disfigurement,lifelong disability,or death.Dermoscopic images acquired in primary-care settings show high intra-class visual similarity and ...Skin diseases affect millions worldwide.Early detection is key to preventing disfigurement,lifelong disability,or death.Dermoscopic images acquired in primary-care settings show high intra-class visual similarity and severe class imbalance,and occasional imaging artifacts can create ambiguity for state-of-the-art convolutional neural networks(CNNs).We frame skin lesion recognition as graph-based reasoning and,to ensure fair evaluation and avoid data leakage,adopt a strict lesion-level partitioning strategy.Each image is first over-segmented using SLIC(Simple Linear Iterative Clustering)to produce perceptually homogeneous superpixels.These superpixels form the nodes of a region-adjacency graph whose edges encode spatial continuity.Node attributes are 1280-dimensional embeddings extracted with a lightweight yet expressive EfficientNet-B0 backbone,providing strong representational power at modest computational cost.The resulting graphs are processed by a five-layer Graph Attention Network(GAT)that learns to weight inter-node relationships dynamically and aggregates multi-hop context before classifying lesions into seven classes with a log-softmax output.Extensive experiments on the DermaMNIST benchmark show the proposed pipeline achieves 88.35%accuracy and 98.04%AUC,outperforming contemporary CNNs,AutoML approaches,and alternative graph neural networks.An ablation study indicates EfficientNet-B0 produces superior node descriptors compared with ResNet-18 and DenseNet,and that roughly five GAT layers strike a good balance between being too shallow and over-deep while avoiding oversmoothing.The method requires no data augmentation or external metadata,making it a drop-in upgrade for clinical computer-aided diagnosis systems.展开更多
Traditional Chinese medicine(TCM)features complex compatibility mechanisms involving multicomponent,multi-target,and multi-pathway interactions.This study presents an interpretable graph artificial intelligence(GraphA...Traditional Chinese medicine(TCM)features complex compatibility mechanisms involving multicomponent,multi-target,and multi-pathway interactions.This study presents an interpretable graph artificial intelligence(GraphAI)framework to quantify such mechanisms in Chinese herbal formulas(CHFs).A multidimensional TCM knowledge graph(TCM-MKG;https://zenodo.org/records/13763953)was constructed,integrating seven standardized modules:TCM terminology,Chinese patent medicines(CPMs),Chinese herbal pieces(CHPs),pharmacognostic origins(POs),chemical compounds,biological targets,and diseases.A neighbor-diffusion strategy was used to address the sparsity of compound-target associations,increasing target coverage from 12.0%to 98.7%.Graph neural networks(GNNs)with attention mechanisms were applied to 6,080 CHFs,modeled as graphs with CHPs as nodes.To embed domain-specific semantics,virtual nodes medicinal properties,i.e.,therapeutic nature,flavor,and meridian tropism,were introduced,enabling interpretable modeling of inter-CHP relationships.The model quantitatively captured classical compatibility roles such as“monarch-minister-assistant-guide”,and uncovered TCM etiological types derived from diagnostic and efficacy patterns.Model validation using 215 CHFs used for coronavirus disease 2019(COVID-19)management highlighted Radix Astragali-Rhizoma Phragmitis as a high-attention herb pair.Mass spectrometry(MS)and target prediction identified three active compounds,i.e.,methylinissolin-3-O-glucoside,corydalin,and pingbeinine,which converge on pathways such as neuroactive ligand-receptor interaction,xenobiotic response,and neuronal function,supporting their neuroimmune and detoxification potential.Given their high safety and dietary compatibility,this herb pair may offer therapeutic value for managing long COVID-19.All data and code are openly available(https://github.com/ZENGJingqi/GraphAI-for-TCM),providing a scalable and interpretable platform for TCM mechanism research and discovery of bioactive herbal constituents.展开更多
Spectrum-based fault localization (SBFL) generates a ranked list of suspicious elements by using the program execution spectrum, but the excessive number of elements ranked in parallel results in low localization accu...Spectrum-based fault localization (SBFL) generates a ranked list of suspicious elements by using the program execution spectrum, but the excessive number of elements ranked in parallel results in low localization accuracy. Most researchers consider intra-class dependencies to improve localization accuracy. However, some studies show that inter-class method call type faults account for more than 20%, which means such methods still have certain limitations. To solve the above problems, this paper proposes a two-phase software fault localization based on relational graph convolutional neural networks (Two-RGCNFL). Firstly, in Phase 1, the method call dependence graph (MCDG) of the program is constructed, the intra-class and inter-class dependencies in MCDG are extracted by using the relational graph convolutional neural network, and the classifier is used to identify the faulty methods. Then, the GraphSMOTE algorithm is improved to alleviate the impact of class imbalance on classification accuracy. Aiming at the problem of parallel ranking of element suspicious values in traditional SBFL technology, in Phase 2, Doc2Vec is used to learn static features, while spectrum information serves as dynamic features. A RankNet model based on siamese multi-layer perceptron is constructed to score and rank statements in the faulty method. This work conducts experiments on 5 real projects of Defects4J benchmark. Experimental results show that, compared with the traditional SBFL technique and two baseline methods, our approach improves the Top-1 accuracy by 262.86%, 29.59% and 53.01%, respectively, which verifies the effectiveness of Two-RGCNFL. Furthermore, this work verifies the importance of inter-class dependencies through ablation experiments.展开更多
Accurate and efficient prediction of the distribution of surface loads on buildings subjected to explosive effects is crucial for rapidly calculating structural dynamic responses,establishing effective protective meas...Accurate and efficient prediction of the distribution of surface loads on buildings subjected to explosive effects is crucial for rapidly calculating structural dynamic responses,establishing effective protective measures,and designing civil defense engineering solutions.Current state-of-the-art methods face several issues:Experimental research is difficult and costly to implement,theoretical research is limited to simple geometries and lacks precision,and direct simulations require substantial computational resources.To address these challenges,this paper presents a data-driven method for predicting blast loads on building surfaces.This approach increases both the accuracy and computational efficiency of load predictions when the geometry of the building changes while the explosive yield remains constant,significantly improving its applicability in complex scenarios.This study introduces an innovative encoder-decoder graph neural network model named BlastGraphNet,which uses a message-passing mechanism to predict the overpressure and impulse load distributions on buildings with conventional and complex geometries during explosive events.The model also facilitates related downstream applications,such as damage mode identification and rapid assessment of virtual city explosions.The calculation results indicate that the prediction error of the model for conventional building tests is less than 2%,and its inference speed is 3-4 orders of magnitude faster than that of state-of-the-art numerical methods.In extreme test cases involving buildings with complex geometries and building clusters,the method achieved high accuracy and excellent generalizability.The strong adaptability and generalizability of BlastGraphNet confirm that this novel method enables precise real-time prediction of blast loads and provides a new paradigm for damage assessment in protective engineering.展开更多
Accurate prediction of molecular properties is crucial for selecting compounds with ideal properties and reducing the costs and risks of trials.Traditional methods based on manually crafted features and graph-based me...Accurate prediction of molecular properties is crucial for selecting compounds with ideal properties and reducing the costs and risks of trials.Traditional methods based on manually crafted features and graph-based methods have shown promising results in molecular property prediction.However,traditional methods rely on expert knowledge and often fail to capture the complex structures and interactions within molecules.Similarly,graph-based methods typically overlook the chemical structure and function hidden in molecular motifs and struggle to effectively integrate global and local molecular information.To address these limitations,we propose a novel fingerprint-enhanced hierarchical graph neural network(FH-GNN)for molecular property prediction that simultaneously learns information from hierarchical molecular graphs and fingerprints.The FH-GNN captures diverse hierarchical chemical information by applying directed message-passing neural networks(D-MPNN)on a hierarchical molecular graph that integrates atomic-level,motif-level,and graph-level information along with their relationships.Addi-tionally,we used an adaptive attention mechanism to balance the importance of hierarchical graphs and fingerprint features,creating a comprehensive molecular embedding that integrated hierarchical mo-lecular structures with domain knowledge.Experiments on eight benchmark datasets from MoleculeNet showed that FH-GNN outperformed the baseline models in both classification and regression tasks for molecular property prediction,validating its capability to comprehensively capture molecular informa-tion.By integrating molecular structure and chemical knowledge,FH-GNN provides a powerful tool for the accurate prediction of molecular properties and aids in the discovery of potential drug candidates.展开更多
Fault diagnosis occupies a pivotal position within the domain of machine and equipment management.Existing methods,however,often exhibit limitations in their scope of application,typically focusing on specific types o...Fault diagnosis occupies a pivotal position within the domain of machine and equipment management.Existing methods,however,often exhibit limitations in their scope of application,typically focusing on specific types of signals or faults in individual mechanical components while being constrained by data types and inherent characteristics.To address the limitations of existing methods,we propose a fault diagnosis method based on graph neural networks(GNNs)embedded with multirelationships of intrinsic mode functions(MIMF).The approach introduces a novel graph topological structure constructed from the features of intrinsic mode functions(IMFs)of monitored signals and their multirelationships.Additionally,a graph-level based fault diagnosis network model is designed to enhance feature learning capabilities for graph samples and enable flexible application across diverse signal sources and devices.Experimental validation with datasets including independent vibration signals for gear fault detection,mixed vibration signals for concurrent gear and bearing faults,and pressure signals for hydraulic cylinder leakage characterization demonstrates the model's adaptability and superior diagnostic accuracy across various types of signals and mechanical systems.展开更多
Schizophrenia(SZ)stands as a severe psychiatric disorder.This study applied diffusion tensor imaging(DTI)data in conjunction with graph neural networks to distinguish SZ patients from normal controls(NCs)and showcases...Schizophrenia(SZ)stands as a severe psychiatric disorder.This study applied diffusion tensor imaging(DTI)data in conjunction with graph neural networks to distinguish SZ patients from normal controls(NCs)and showcases the superior performance of a graph neural network integrating combined fractional anisotropy and fiber number brain network features,achieving an accuracy of 73.79%in distinguishing SZ patients from NCs.Beyond mere discrimination,our study delved deeper into the advantages of utilizing white matter brain network features for identifying SZ patients through interpretable model analysis and gene expression analysis.These analyses uncovered intricate interrelationships between brain imaging markers and genetic biomarkers,providing novel insights into the neuropathological basis of SZ.In summary,our findings underscore the potential of graph neural networks applied to multimodal DTI data for enhancing SZ detection through an integrated analysis of neuroimaging and genetic features.展开更多
Recently,numerous estimation issues have been solved due to the developments in data-driven artificial neural networks(ANN)and graph neural networks(GNN).The primary limitation of previous methodologies has been the d...Recently,numerous estimation issues have been solved due to the developments in data-driven artificial neural networks(ANN)and graph neural networks(GNN).The primary limitation of previous methodologies has been the dependence on data that can be structured in a grid format.However,physiological recordings often exhibit irregular and unordered patterns,posing a significant challenge in conceptualising them as matrices.As a result,GNNs which comprise interactive nodes connected by edges whose weights are defined by anatomical junctions or temporal relationships have received a lot of consideration by leveraging implicit data that exists in a biological system.Additionally,our study incorporates a structural GNN to effectively differentiate between different degrees of infection in both the left and right hemispheres of the brain.Subsequently,demographic data are included,and a multi-task learning architecture is devised,integrating classification and regression tasks.The trials used an authentic dataset,including 800 brain x-ray pictures,consisting of 560 instances classified as moderate cases and 240 instances classified as severe cases.Based on empirical evidence,our methodology demonstrates superior performance in classification,surpassing other comparison methods with a notable achievement of 92.27%in terms of area under the curve as well as a correlation coefficient of 0.62.展开更多
The primary goal of software defect prediction (SDP) is to pinpoint code modules that are likely to contain defects, thereby enabling software quality assurance teams to strategically allocate their resources and manp...The primary goal of software defect prediction (SDP) is to pinpoint code modules that are likely to contain defects, thereby enabling software quality assurance teams to strategically allocate their resources and manpower. Within-project defect prediction (WPDP) is a widely used method in SDP. Despite various improvements, current methods still face challenges such as coarse-grained prediction and ineffective handling of data drift due to differences in project distribution. To address these issues, we propose a fine-grained SDP method called DIDP (drift-immune defect prediction), based on drift-immune graph neural networks (DI-GNN). DIDP converts source code into graph representations and uses DI-GNN to mitigate data drift at the model level. It also analyses key statements leading to file defects for a more detailed SDP approach. We evaluated the performance of DIDP in WPDP by examining its file-level and statement-level accuracy compared to state-of-the-art methods, and by examining its cross-project prediction accuracy. The results of the experiment show that DIDP showed significant improvements in F1-score and Recall@Top20%LOC compared to existing methods, even with large software version changes. DIDP also performed well in cross-project SDP. Our study demonstrates that DIDP achieves impressive prediction results in WPDP, effectively mitigating data drift and accurately predicting defective files. Additionally, DIDP can rank the risk of statements in defective files, aiding developers and testers in identifying potential code issues.展开更多
The increasing frequency of offshore engineering activities,particularly the expansion of offshore oil transport and the rise in the number of oil platforms,has greatly increased the potential risk of marine oil spill...The increasing frequency of offshore engineering activities,particularly the expansion of offshore oil transport and the rise in the number of oil platforms,has greatly increased the potential risk of marine oil spill incidents.Historically,several large oil spills have had long-term adverse effects on marine ecosystems and economic development,highlighting the importance of accurate-ly delineating and monitoring oil spill areas.In this study,graph neural network technology is introduced to implement semantic seg-mentation of SAR images,and two graph neural network models based on Graph-FCN and Graph-DeepLabV3+with the introduction of an attention mechanism are constructed and evaluated to improve the accuracy and efficiency of oil spill detection.By com-paring the Swin-Unet model,the Graph-DeepLabV3+model performs better in complex scenarios,especially in edge detail recognition.This not only provides strong technical support for marine oil spill monitoring but also provides an effective solution to deal with the potential risks brought by the increase of marine engineering activities,which is of great practical significance as it helps to safeguard the health and sustainable development of marine ecosystems and reduce the economic losses.展开更多
Weather forecasting is crucial for agriculture,transportation,and industry.Deep Learning(DL)has greatly improved the prediction accuracy.Among them,Graph Neural Networks(GNNs)excel at processing weather data by establ...Weather forecasting is crucial for agriculture,transportation,and industry.Deep Learning(DL)has greatly improved the prediction accuracy.Among them,Graph Neural Networks(GNNs)excel at processing weather data by establishing connections between regions.This allows them to understand complex patterns that traditional methods might miss.As a result,achieving more accurate predictions becomes possible.The paper reviews the role of GNNs in short-to medium-range weather forecasting.The methods are classified into three categories based on dataset differences.The paper also further identifies five promising research frontiers.These areas aim to boost forecasting precision and enhance computational efficiency.They offer valuable insights for future weather forecasting systems.展开更多
With the advancement of Vehicle-to-Everything(V2X)technology,efficient resource allocation in dynamic vehicular networks has become a critical challenge for achieving optimal performance.Existing methods suffer from h...With the advancement of Vehicle-to-Everything(V2X)technology,efficient resource allocation in dynamic vehicular networks has become a critical challenge for achieving optimal performance.Existing methods suffer from high computational complexity and decision latency under high-density traffic and heterogeneous network conditions.To address these challenges,this study presents an innovative framework that combines Graph Neural Networks(GNNs)with a Double Deep Q-Network(DDQN),utilizing dynamic graph structures and reinforcement learning.An adaptive neighbor sampling mechanism is introduced to dynamically select the most relevant neighbors based on interference levels and network topology,thereby improving decision accuracy and efficiency.Meanwhile,the framework models communication links as nodes and interference relationships as edges,effectively capturing the direct impact of interference on resource allocation while reducing computational complexity and preserving critical interaction information.Employing an aggregation mechanism based on the Graph Attention Network(GAT),it dynamically adjusts the neighbor sampling scope and performs attention-weighted aggregation based on node importance,ensuring more efficient and adaptive resource management.This design ensures reliable Vehicle-to-Vehicle(V2V)communication while maintaining high Vehicle-to-Infrastructure(V2I)throughput.The framework retains the global feature learning capabilities of GNNs and supports distributed network deployment,allowing vehicles to extract low-dimensional graph embeddings from local observations for real-time resource decisions.Experimental results demonstrate that the proposed method significantly reduces computational overhead,mitigates latency,and improves resource utilization efficiency in vehicular networks under complex traffic scenarios.This research not only provides a novel solution to resource allocation challenges in V2X networks but also advances the application of DDQN in intelligent transportation systems,offering substantial theoretical significance and practical value.展开更多
Recent advances in statistical physics highlight the significant potential of machine learning for phase transition recognition.This study introduces a deep learning framework based on graph neural network to investig...Recent advances in statistical physics highlight the significant potential of machine learning for phase transition recognition.This study introduces a deep learning framework based on graph neural network to investigate non-equilibrium phase transitions,specifically focusing on the directed percolation process.By converting lattices with varying dimensions and connectivity schemes into graph structures and embedding the temporal evolution of the percolation process into node features,our approach enables unified analysis across diverse systems.The framework utilizes a multi-layer graph attention mechanism combined with global pooling to autonomously extract critical features from local dynamics to global phase transition signatures.The model successfully predicts percolation thresholds without relying on lattice geometry,demonstrating its robustness and versatility.Our approach not only offers new insights into phase transition studies but also provides a powerful tool for analyzing complex dynamical systems across various domains.展开更多
Studies show that Graph Neural Networks(GNNs)are susceptible to minor perturbations.Therefore,analyzing adversarial attacks on GNNs is crucial in current research.Previous studies used Generative Adversarial Networks ...Studies show that Graph Neural Networks(GNNs)are susceptible to minor perturbations.Therefore,analyzing adversarial attacks on GNNs is crucial in current research.Previous studies used Generative Adversarial Networks to generate a set of fake nodes,injecting them into a clean GNNs to poison the graph structure and evaluate the robustness of GNNs.In the attack process,the computation of new node connections and the attack loss are independent,which affects the attack on the GNN.To improve this,a Fake Node Camouflage Attack based on Mutual Information(FNCAMI)algorithm is proposed.By incorporating Mutual Information(MI)loss,the distribution of nodes injected into the GNNs become more similar to the original nodes,achieving better attack results.Since the loss ratios of GNNs and MI affect performance,we also design an adaptive weighting method.By adjusting the loss weights in real-time through rate changes,larger loss values are obtained,eliminating local optima.The feasibility,effectiveness,and stealthiness of this algorithm are validated on four real datasets.Additionally,we use both global and targeted attacks to test the algorithm’s performance.Comparisons with baseline attack algorithms and ablation experiments demonstrate the efficiency of the FNCAMI algorithm.展开更多
With the emphasis on user privacy and communication security, encrypted traffic has increased dramatically, which brings great challenges to traffic classification. The classification method of encrypted traffic based...With the emphasis on user privacy and communication security, encrypted traffic has increased dramatically, which brings great challenges to traffic classification. The classification method of encrypted traffic based on GNN can deal with encrypted traffic well. However, existing GNN-based approaches ignore the relationship between client or server packets. In this paper, we design a network traffic topology based on GCN, called Flow Mapping Graph (FMG). FMG establishes sequential edges between vertexes by the arrival order of packets and establishes jump-order edges between vertexes by connecting packets in different bursts with the same direction. It not only reflects the time characteristics of the packet but also strengthens the relationship between the client or server packets. According to FMG, a Traffic Mapping Classification model (TMC-GCN) is designed, which can automatically capture and learn the characteristics and structure information of the top vertex in FMG. The TMC-GCN model is used to classify the encrypted traffic. The encryption stream classification problem is transformed into a graph classification problem, which can effectively deal with data from different data sources and application scenarios. By comparing the performance of TMC-GCN with other classical models in four public datasets, including CICIOT2023, ISCXVPN2016, CICAAGM2017, and GraphDapp, the effectiveness of the FMG algorithm is verified. The experimental results show that the accuracy rate of the TMC-GCN model is 96.13%, the recall rate is 95.04%, and the F1 rate is 94.54%.展开更多
The ability to accurately predict urban traffic flows is crucial for optimising city operations.Consequently,various methods for forecasting urban traffic have been developed,focusing on analysing historical data to u...The ability to accurately predict urban traffic flows is crucial for optimising city operations.Consequently,various methods for forecasting urban traffic have been developed,focusing on analysing historical data to understand complex mobility patterns.Deep learning techniques,such as graph neural networks(GNNs),are popular for their ability to capture spatio-temporal dependencies.However,these models often become overly complex due to the large number of hyper-parameters involved.In this study,we introduce Dynamic Multi-Graph Spatial-Temporal Graph Neural Ordinary Differential Equation Networks(DMST-GNODE),a framework based on ordinary differential equations(ODEs)that autonomously discovers effective spatial-temporal graph neural network(STGNN)architectures for traffic prediction tasks.The comparative analysis of DMST-GNODE and baseline models indicates that DMST-GNODE model demonstrates superior performance across multiple datasets,consistently achieving the lowest Root Mean Square Error(RMSE)and Mean Absolute Error(MAE)values,alongside the highest accuracy.On the BKK(Bangkok)dataset,it outperformed other models with an RMSE of 3.3165 and an accuracy of 0.9367 for a 20-min interval,maintaining this trend across 40 and 60 min.Similarly,on the PeMS08 dataset,DMST-GNODE achieved the best performance with an RMSE of 19.4863 and an accuracy of 0.9377 at 20 min,demonstrating its effectiveness over longer periods.The Los_Loop dataset results further emphasise this model’s advantage,with an RMSE of 3.3422 and an accuracy of 0.7643 at 20 min,consistently maintaining superiority across all time intervals.These numerical highlights indicate that DMST-GNODE not only outperforms baseline models but also achieves higher accuracy and lower errors across different time intervals and datasets.展开更多
基金supported by the National Natural Science Foundation of China(Grant No.12272316).
文摘The adjoint method is widely used in gradient-based optimization with high-dimensional design variables.However,the cost of solving the adjoint equations in each iteration is comparable to that of solving the flow field,resulting in expensive computational costs.To improve the efficiency of solving adjoint equations,we propose a physics-constrained graph neural networks for solving adjoint equations,named ADJ-PCGN.ADJ-PCGN establishes a mapping relationship between flow characteristics and adjoint vector based on data,serving as a replacement for the computationally expensive numerical solution of adjoint equations.A physics-based graph structure and message-passing mechanism are designed to endow its strong fitting and generalization capabilities.Taking transonic drag reduction and maximum lift-drag ratio of the airfoil as examples,results indicate that ADJ-PCGN attains a similar optimal shape as the classical direct adjoint loop method.In addition,ADJ-PCGN demonstrates strong generalization capabilities across different mesh topologies,mesh densities,and out-of-distribution conditions.It holds the potential to become a universal model for aerodynamic shape optimization involving states,geometries,and meshes.
文摘The increasing popularity of the Internet and the widespread use of information technology have led to a rise in the number and sophistication of network attacks and security threats.Intrusion detection systems are crucial to network security,playing a pivotal role in safeguarding networks from potential threats.However,in the context of an evolving landscape of sophisticated and elusive attacks,existing intrusion detection methodologies often overlook critical aspects such as changes in network topology over time and interactions between hosts.To address these issues,this paper proposes a real-time network intrusion detection method based on graph neural networks.The proposedmethod leverages the advantages of graph neural networks and employs a straightforward graph construction method to represent network traffic as dynamic graph-structured data.Additionally,a graph convolution operation with a multi-head attention mechanism is utilized to enhance the model’s ability to capture the intricate relationships within the graph structure comprehensively.Furthermore,it uses an integrated graph neural network to address dynamic graphs’structural and topological changes at different time points and the challenges of edge embedding in intrusion detection data.The edge classification problem is effectively transformed into node classification by employing a line graph data representation,which facilitates fine-grained intrusion detection tasks on dynamic graph node feature representations.The efficacy of the proposed method is evaluated using two commonly used intrusion detection datasets,UNSW-NB15 and NF-ToN-IoT-v2,and results are compared with previous studies in this field.The experimental results demonstrate that our proposed method achieves 99.3%and 99.96%accuracy on the two datasets,respectively,and outperforms the benchmark model in several evaluation metrics.
基金supported by the National Natural Science Foundation of China(Nos.62176122 and 62061146002).
文摘Federated Graph Neural Networks (FedGNNs) have achieved significant success in representation learning for graph data, enabling collaborative training among multiple parties without sharing their raw graph data and solving the data isolation problem faced by centralized GNNs in data-sensitive scenarios. Despite the plethora of prior work on inference attacks against centralized GNNs, the vulnerability of FedGNNs to inference attacks has not yet been widely explored. It is still unclear whether the privacy leakage risks of centralized GNNs will also be introduced in FedGNNs. To bridge this gap, we present PIAFGNN, the first property inference attack (PIA) against FedGNNs. Compared with prior works on centralized GNNs, in PIAFGNN, the attacker can only obtain the global embedding gradient distributed by the central server. The attacker converts the task of stealing the target user’s local embeddings into a regression problem, using a regression model to generate the target graph node embeddings. By training shadow models and property classifiers, the attacker can infer the basic property information within the target graph that is of interest. Experiments on three benchmark graph datasets demonstrate that PIAFGNN achieves attack accuracy of over 70% in most cases, even approaching the attack accuracy of inference attacks against centralized GNNs in some instances, which is much higher than the attack accuracy of the random guessing method. Furthermore, we observe that common defense mechanisms cannot mitigate our attack without affecting the model’s performance on mainly classification tasks.
基金funded by the Key Research and Development Program of Zhejiang Province No.2023C01141the Science and Technology Innovation Community Project of the Yangtze River Delta No.23002410100suported by the Open Research Fund of the State Key Laboratory of Blockchain and Data Security,Zhejiang University.
文摘Graph Neural Networks(GNNs)have demonstrated outstanding capabilities in processing graph-structured data and are increasingly being integrated into large-scale pre-trained models,such as Large Language Models(LLMs),to enhance structural reasoning,knowledge retrieval,and memory management.The expansion of their application scope imposes higher requirements on the robustness of GNNs.However,as GNNs are applied to more dynamic and heterogeneous environments,they become increasingly vulnerable to real-world perturbations.In particular,graph data frequently encounters joint adversarial perturbations that simultaneously affect both structures and features,which are significantly more challenging than isolated attacks.These disruptions,caused by incomplete data,malicious attacks,or inherent noise,pose substantial threats to the stable and reliable performance of traditional GNN models.To address this issue,this study proposes the Dual-Shield Graph Neural Network(DSGNN),a defense model that simultaneously mitigates structural and feature perturbations.DSGNN utilizes two parallel GNN channels to independently process structural noise and feature noise,and introduces an adaptive fusion mechanism that integrates information from both pathways to generate robust node representations.Theoretical analysis demonstrates that DSGNN achieves a tighter robustness boundary under joint perturbations compared to conventional single-channel methods.Experimental evaluations across Cora,CiteSeer,and Industry datasets show that DSGNN achieves the highest average classification accuracy under various adversarial settings,reaching 81.24%,71.94%,and 81.66%,respectively,outperforming GNNGuard,GCN-Jaccard,GCN-SVD,RGCN,and NoisyGNN.These results underscore the importance of multi-view perturbation decoupling in constructing resilient GNN models for real-world applications.
基金funded by the Deanship of Graduate Studies and Scientific Research at Jouf University under grant No.(DGSSR-2025-02-01296).
文摘Skin diseases affect millions worldwide.Early detection is key to preventing disfigurement,lifelong disability,or death.Dermoscopic images acquired in primary-care settings show high intra-class visual similarity and severe class imbalance,and occasional imaging artifacts can create ambiguity for state-of-the-art convolutional neural networks(CNNs).We frame skin lesion recognition as graph-based reasoning and,to ensure fair evaluation and avoid data leakage,adopt a strict lesion-level partitioning strategy.Each image is first over-segmented using SLIC(Simple Linear Iterative Clustering)to produce perceptually homogeneous superpixels.These superpixels form the nodes of a region-adjacency graph whose edges encode spatial continuity.Node attributes are 1280-dimensional embeddings extracted with a lightweight yet expressive EfficientNet-B0 backbone,providing strong representational power at modest computational cost.The resulting graphs are processed by a five-layer Graph Attention Network(GAT)that learns to weight inter-node relationships dynamically and aggregates multi-hop context before classifying lesions into seven classes with a log-softmax output.Extensive experiments on the DermaMNIST benchmark show the proposed pipeline achieves 88.35%accuracy and 98.04%AUC,outperforming contemporary CNNs,AutoML approaches,and alternative graph neural networks.An ablation study indicates EfficientNet-B0 produces superior node descriptors compared with ResNet-18 and DenseNet,and that roughly five GAT layers strike a good balance between being too shallow and over-deep while avoiding oversmoothing.The method requires no data augmentation or external metadata,making it a drop-in upgrade for clinical computer-aided diagnosis systems.
基金supported by the National Natural Science Foundation of China(Grant No.:82230117).
文摘Traditional Chinese medicine(TCM)features complex compatibility mechanisms involving multicomponent,multi-target,and multi-pathway interactions.This study presents an interpretable graph artificial intelligence(GraphAI)framework to quantify such mechanisms in Chinese herbal formulas(CHFs).A multidimensional TCM knowledge graph(TCM-MKG;https://zenodo.org/records/13763953)was constructed,integrating seven standardized modules:TCM terminology,Chinese patent medicines(CPMs),Chinese herbal pieces(CHPs),pharmacognostic origins(POs),chemical compounds,biological targets,and diseases.A neighbor-diffusion strategy was used to address the sparsity of compound-target associations,increasing target coverage from 12.0%to 98.7%.Graph neural networks(GNNs)with attention mechanisms were applied to 6,080 CHFs,modeled as graphs with CHPs as nodes.To embed domain-specific semantics,virtual nodes medicinal properties,i.e.,therapeutic nature,flavor,and meridian tropism,were introduced,enabling interpretable modeling of inter-CHP relationships.The model quantitatively captured classical compatibility roles such as“monarch-minister-assistant-guide”,and uncovered TCM etiological types derived from diagnostic and efficacy patterns.Model validation using 215 CHFs used for coronavirus disease 2019(COVID-19)management highlighted Radix Astragali-Rhizoma Phragmitis as a high-attention herb pair.Mass spectrometry(MS)and target prediction identified three active compounds,i.e.,methylinissolin-3-O-glucoside,corydalin,and pingbeinine,which converge on pathways such as neuroactive ligand-receptor interaction,xenobiotic response,and neuronal function,supporting their neuroimmune and detoxification potential.Given their high safety and dietary compatibility,this herb pair may offer therapeutic value for managing long COVID-19.All data and code are openly available(https://github.com/ZENGJingqi/GraphAI-for-TCM),providing a scalable and interpretable platform for TCM mechanism research and discovery of bioactive herbal constituents.
基金funded by the Youth Fund of the National Natural Science Foundation of China(Grant No.42261070).
文摘Spectrum-based fault localization (SBFL) generates a ranked list of suspicious elements by using the program execution spectrum, but the excessive number of elements ranked in parallel results in low localization accuracy. Most researchers consider intra-class dependencies to improve localization accuracy. However, some studies show that inter-class method call type faults account for more than 20%, which means such methods still have certain limitations. To solve the above problems, this paper proposes a two-phase software fault localization based on relational graph convolutional neural networks (Two-RGCNFL). Firstly, in Phase 1, the method call dependence graph (MCDG) of the program is constructed, the intra-class and inter-class dependencies in MCDG are extracted by using the relational graph convolutional neural network, and the classifier is used to identify the faulty methods. Then, the GraphSMOTE algorithm is improved to alleviate the impact of class imbalance on classification accuracy. Aiming at the problem of parallel ranking of element suspicious values in traditional SBFL technology, in Phase 2, Doc2Vec is used to learn static features, while spectrum information serves as dynamic features. A RankNet model based on siamese multi-layer perceptron is constructed to score and rank statements in the faulty method. This work conducts experiments on 5 real projects of Defects4J benchmark. Experimental results show that, compared with the traditional SBFL technique and two baseline methods, our approach improves the Top-1 accuracy by 262.86%, 29.59% and 53.01%, respectively, which verifies the effectiveness of Two-RGCNFL. Furthermore, this work verifies the importance of inter-class dependencies through ablation experiments.
基金supported by the National Natural Science Founion of China(U2241285).
文摘Accurate and efficient prediction of the distribution of surface loads on buildings subjected to explosive effects is crucial for rapidly calculating structural dynamic responses,establishing effective protective measures,and designing civil defense engineering solutions.Current state-of-the-art methods face several issues:Experimental research is difficult and costly to implement,theoretical research is limited to simple geometries and lacks precision,and direct simulations require substantial computational resources.To address these challenges,this paper presents a data-driven method for predicting blast loads on building surfaces.This approach increases both the accuracy and computational efficiency of load predictions when the geometry of the building changes while the explosive yield remains constant,significantly improving its applicability in complex scenarios.This study introduces an innovative encoder-decoder graph neural network model named BlastGraphNet,which uses a message-passing mechanism to predict the overpressure and impulse load distributions on buildings with conventional and complex geometries during explosive events.The model also facilitates related downstream applications,such as damage mode identification and rapid assessment of virtual city explosions.The calculation results indicate that the prediction error of the model for conventional building tests is less than 2%,and its inference speed is 3-4 orders of magnitude faster than that of state-of-the-art numerical methods.In extreme test cases involving buildings with complex geometries and building clusters,the method achieved high accuracy and excellent generalizability.The strong adaptability and generalizability of BlastGraphNet confirm that this novel method enables precise real-time prediction of blast loads and provides a new paradigm for damage assessment in protective engineering.
基金supported by Macao Science and Technology Development Fund,Macao SAR,China(Grant No.:0043/2023/AFJ)the National Natural Science Foundation of China(Grant No.:22173038)Macao Polytechnic University,Macao SAR,China(Grant No.:RP/FCA-01/2022).
文摘Accurate prediction of molecular properties is crucial for selecting compounds with ideal properties and reducing the costs and risks of trials.Traditional methods based on manually crafted features and graph-based methods have shown promising results in molecular property prediction.However,traditional methods rely on expert knowledge and often fail to capture the complex structures and interactions within molecules.Similarly,graph-based methods typically overlook the chemical structure and function hidden in molecular motifs and struggle to effectively integrate global and local molecular information.To address these limitations,we propose a novel fingerprint-enhanced hierarchical graph neural network(FH-GNN)for molecular property prediction that simultaneously learns information from hierarchical molecular graphs and fingerprints.The FH-GNN captures diverse hierarchical chemical information by applying directed message-passing neural networks(D-MPNN)on a hierarchical molecular graph that integrates atomic-level,motif-level,and graph-level information along with their relationships.Addi-tionally,we used an adaptive attention mechanism to balance the importance of hierarchical graphs and fingerprint features,creating a comprehensive molecular embedding that integrated hierarchical mo-lecular structures with domain knowledge.Experiments on eight benchmark datasets from MoleculeNet showed that FH-GNN outperformed the baseline models in both classification and regression tasks for molecular property prediction,validating its capability to comprehensively capture molecular informa-tion.By integrating molecular structure and chemical knowledge,FH-GNN provides a powerful tool for the accurate prediction of molecular properties and aids in the discovery of potential drug candidates.
文摘Fault diagnosis occupies a pivotal position within the domain of machine and equipment management.Existing methods,however,often exhibit limitations in their scope of application,typically focusing on specific types of signals or faults in individual mechanical components while being constrained by data types and inherent characteristics.To address the limitations of existing methods,we propose a fault diagnosis method based on graph neural networks(GNNs)embedded with multirelationships of intrinsic mode functions(MIMF).The approach introduces a novel graph topological structure constructed from the features of intrinsic mode functions(IMFs)of monitored signals and their multirelationships.Additionally,a graph-level based fault diagnosis network model is designed to enhance feature learning capabilities for graph samples and enable flexible application across diverse signal sources and devices.Experimental validation with datasets including independent vibration signals for gear fault detection,mixed vibration signals for concurrent gear and bearing faults,and pressure signals for hydraulic cylinder leakage characterization demonstrates the model's adaptability and superior diagnostic accuracy across various types of signals and mechanical systems.
基金supported by the National Natural Science Foundation of China(62276049,61701078,61872068,and 62006038)the Natural Science Foundation of Sichuan Province(2025ZNSFSC0487)+3 种基金the Science and Technology Innovation 2030-Brain Science and Brain-Inspired Intelligence Project(2021ZD0200200)the National Key R&D Program of China(2023YFE0118600)Sichuan Province Science and Technology Support Program(2019YJ0193,2021YFG0126,2021YFG0366,and 2022YFS0180)Medico-Engineering Cooperation Funds from the University of Electronic Science and Technology of China(ZYGX2021YGLH014).
文摘Schizophrenia(SZ)stands as a severe psychiatric disorder.This study applied diffusion tensor imaging(DTI)data in conjunction with graph neural networks to distinguish SZ patients from normal controls(NCs)and showcases the superior performance of a graph neural network integrating combined fractional anisotropy and fiber number brain network features,achieving an accuracy of 73.79%in distinguishing SZ patients from NCs.Beyond mere discrimination,our study delved deeper into the advantages of utilizing white matter brain network features for identifying SZ patients through interpretable model analysis and gene expression analysis.These analyses uncovered intricate interrelationships between brain imaging markers and genetic biomarkers,providing novel insights into the neuropathological basis of SZ.In summary,our findings underscore the potential of graph neural networks applied to multimodal DTI data for enhancing SZ detection through an integrated analysis of neuroimaging and genetic features.
文摘Recently,numerous estimation issues have been solved due to the developments in data-driven artificial neural networks(ANN)and graph neural networks(GNN).The primary limitation of previous methodologies has been the dependence on data that can be structured in a grid format.However,physiological recordings often exhibit irregular and unordered patterns,posing a significant challenge in conceptualising them as matrices.As a result,GNNs which comprise interactive nodes connected by edges whose weights are defined by anatomical junctions or temporal relationships have received a lot of consideration by leveraging implicit data that exists in a biological system.Additionally,our study incorporates a structural GNN to effectively differentiate between different degrees of infection in both the left and right hemispheres of the brain.Subsequently,demographic data are included,and a multi-task learning architecture is devised,integrating classification and regression tasks.The trials used an authentic dataset,including 800 brain x-ray pictures,consisting of 560 instances classified as moderate cases and 240 instances classified as severe cases.Based on empirical evidence,our methodology demonstrates superior performance in classification,surpassing other comparison methods with a notable achievement of 92.27%in terms of area under the curve as well as a correlation coefficient of 0.62.
基金The authors would like to express appreciation to the National Natural Science Foundation of China(Grant No.61762067)for their financial support.
文摘The primary goal of software defect prediction (SDP) is to pinpoint code modules that are likely to contain defects, thereby enabling software quality assurance teams to strategically allocate their resources and manpower. Within-project defect prediction (WPDP) is a widely used method in SDP. Despite various improvements, current methods still face challenges such as coarse-grained prediction and ineffective handling of data drift due to differences in project distribution. To address these issues, we propose a fine-grained SDP method called DIDP (drift-immune defect prediction), based on drift-immune graph neural networks (DI-GNN). DIDP converts source code into graph representations and uses DI-GNN to mitigate data drift at the model level. It also analyses key statements leading to file defects for a more detailed SDP approach. We evaluated the performance of DIDP in WPDP by examining its file-level and statement-level accuracy compared to state-of-the-art methods, and by examining its cross-project prediction accuracy. The results of the experiment show that DIDP showed significant improvements in F1-score and Recall@Top20%LOC compared to existing methods, even with large software version changes. DIDP also performed well in cross-project SDP. Our study demonstrates that DIDP achieves impressive prediction results in WPDP, effectively mitigating data drift and accurately predicting defective files. Additionally, DIDP can rank the risk of statements in defective files, aiding developers and testers in identifying potential code issues.
基金supported by the Natural Science Foun-dation of Shandong Province,China(No.ZR2024QF057)the Natural Science Foundation of Jiangsu Province,China(No.BK20240937)+1 种基金the Natural Science Foundation of China(No.42276215)the China University of Mining and Technology(CUMT)Open Sharing Fund for Large-Scale Instruments and Equipment(No.DYGX-2024-86).
文摘The increasing frequency of offshore engineering activities,particularly the expansion of offshore oil transport and the rise in the number of oil platforms,has greatly increased the potential risk of marine oil spill incidents.Historically,several large oil spills have had long-term adverse effects on marine ecosystems and economic development,highlighting the importance of accurate-ly delineating and monitoring oil spill areas.In this study,graph neural network technology is introduced to implement semantic seg-mentation of SAR images,and two graph neural network models based on Graph-FCN and Graph-DeepLabV3+with the introduction of an attention mechanism are constructed and evaluated to improve the accuracy and efficiency of oil spill detection.By com-paring the Swin-Unet model,the Graph-DeepLabV3+model performs better in complex scenarios,especially in edge detail recognition.This not only provides strong technical support for marine oil spill monitoring but also provides an effective solution to deal with the potential risks brought by the increase of marine engineering activities,which is of great practical significance as it helps to safeguard the health and sustainable development of marine ecosystems and reduce the economic losses.
基金supported by Key Laboratory of Smart Earth(KF2023ZD03-05)CMA Innovative and Development Program(CXFZ.20231035)+2 种基金National Key R&D Program of China(No.2021ZD0111902)National Natural Science Foundation of China(Nos.62472014,U21B2038)the Scientific and Technological Project of China Meteorological Administration(CMAJBGS202505).
文摘Weather forecasting is crucial for agriculture,transportation,and industry.Deep Learning(DL)has greatly improved the prediction accuracy.Among them,Graph Neural Networks(GNNs)excel at processing weather data by establishing connections between regions.This allows them to understand complex patterns that traditional methods might miss.As a result,achieving more accurate predictions becomes possible.The paper reviews the role of GNNs in short-to medium-range weather forecasting.The methods are classified into three categories based on dataset differences.The paper also further identifies five promising research frontiers.These areas aim to boost forecasting precision and enhance computational efficiency.They offer valuable insights for future weather forecasting systems.
基金Project ZR2023MF111 supported by Shandong Provincial Natural Science Foundation。
文摘With the advancement of Vehicle-to-Everything(V2X)technology,efficient resource allocation in dynamic vehicular networks has become a critical challenge for achieving optimal performance.Existing methods suffer from high computational complexity and decision latency under high-density traffic and heterogeneous network conditions.To address these challenges,this study presents an innovative framework that combines Graph Neural Networks(GNNs)with a Double Deep Q-Network(DDQN),utilizing dynamic graph structures and reinforcement learning.An adaptive neighbor sampling mechanism is introduced to dynamically select the most relevant neighbors based on interference levels and network topology,thereby improving decision accuracy and efficiency.Meanwhile,the framework models communication links as nodes and interference relationships as edges,effectively capturing the direct impact of interference on resource allocation while reducing computational complexity and preserving critical interaction information.Employing an aggregation mechanism based on the Graph Attention Network(GAT),it dynamically adjusts the neighbor sampling scope and performs attention-weighted aggregation based on node importance,ensuring more efficient and adaptive resource management.This design ensures reliable Vehicle-to-Vehicle(V2V)communication while maintaining high Vehicle-to-Infrastructure(V2I)throughput.The framework retains the global feature learning capabilities of GNNs and supports distributed network deployment,allowing vehicles to extract low-dimensional graph embeddings from local observations for real-time resource decisions.Experimental results demonstrate that the proposed method significantly reduces computational overhead,mitigates latency,and improves resource utilization efficiency in vehicular networks under complex traffic scenarios.This research not only provides a novel solution to resource allocation challenges in V2X networks but also advances the application of DDQN in intelligent transportation systems,offering substantial theoretical significance and practical value.
基金supported by the Fund from the Science and Technology Department of Henan Province,China(Grant Nos.222102210233 and 232102210064)the National Natural Science Foundation of China(Grant Nos.62373169 and 72474086)+5 种基金the Young and Midcareer Academic Leader of Jiangsu Province,China(Grant No.Qinglan Project in 2024)the National Statistical Science Research Project(Grant No.2022LZ03)Shaanxi Provincial Soft Science Project(Grant No.2022KRM111)Shaanxi Provincial Social Science Foundation(Grant No.2022R016)the Special Project for Philosophical and Social Sciences Research in Shaanxi Province,China(Grant No.2024QN018)the Fund from the Henan Office of Philosophy and Social Science(Grant No.2023CJJ112).
文摘Recent advances in statistical physics highlight the significant potential of machine learning for phase transition recognition.This study introduces a deep learning framework based on graph neural network to investigate non-equilibrium phase transitions,specifically focusing on the directed percolation process.By converting lattices with varying dimensions and connectivity schemes into graph structures and embedding the temporal evolution of the percolation process into node features,our approach enables unified analysis across diverse systems.The framework utilizes a multi-layer graph attention mechanism combined with global pooling to autonomously extract critical features from local dynamics to global phase transition signatures.The model successfully predicts percolation thresholds without relying on lattice geometry,demonstrating its robustness and versatility.Our approach not only offers new insights into phase transition studies but also provides a powerful tool for analyzing complex dynamical systems across various domains.
基金supported by the Natural Science Basic Research Plan in Shaanxi Province of China(Program No.2022JM-381,2017JQ6070)National Natural Science Foundation of China(Grant No.61703256),Foundation of State Key Laboratory of Public Big Data(No.PBD2022-08)the Fundamental Research Funds for the Central Universities,China(Program No.GK202201014,GK202202003,GK201803020).
文摘Studies show that Graph Neural Networks(GNNs)are susceptible to minor perturbations.Therefore,analyzing adversarial attacks on GNNs is crucial in current research.Previous studies used Generative Adversarial Networks to generate a set of fake nodes,injecting them into a clean GNNs to poison the graph structure and evaluate the robustness of GNNs.In the attack process,the computation of new node connections and the attack loss are independent,which affects the attack on the GNN.To improve this,a Fake Node Camouflage Attack based on Mutual Information(FNCAMI)algorithm is proposed.By incorporating Mutual Information(MI)loss,the distribution of nodes injected into the GNNs become more similar to the original nodes,achieving better attack results.Since the loss ratios of GNNs and MI affect performance,we also design an adaptive weighting method.By adjusting the loss weights in real-time through rate changes,larger loss values are obtained,eliminating local optima.The feasibility,effectiveness,and stealthiness of this algorithm are validated on four real datasets.Additionally,we use both global and targeted attacks to test the algorithm’s performance.Comparisons with baseline attack algorithms and ablation experiments demonstrate the efficiency of the FNCAMI algorithm.
基金supported by the National Key Research and Development Program of China No.2023YFA1009500.
文摘With the emphasis on user privacy and communication security, encrypted traffic has increased dramatically, which brings great challenges to traffic classification. The classification method of encrypted traffic based on GNN can deal with encrypted traffic well. However, existing GNN-based approaches ignore the relationship between client or server packets. In this paper, we design a network traffic topology based on GCN, called Flow Mapping Graph (FMG). FMG establishes sequential edges between vertexes by the arrival order of packets and establishes jump-order edges between vertexes by connecting packets in different bursts with the same direction. It not only reflects the time characteristics of the packet but also strengthens the relationship between the client or server packets. According to FMG, a Traffic Mapping Classification model (TMC-GCN) is designed, which can automatically capture and learn the characteristics and structure information of the top vertex in FMG. The TMC-GCN model is used to classify the encrypted traffic. The encryption stream classification problem is transformed into a graph classification problem, which can effectively deal with data from different data sources and application scenarios. By comparing the performance of TMC-GCN with other classical models in four public datasets, including CICIOT2023, ISCXVPN2016, CICAAGM2017, and GraphDapp, the effectiveness of the FMG algorithm is verified. The experimental results show that the accuracy rate of the TMC-GCN model is 96.13%, the recall rate is 95.04%, and the F1 rate is 94.54%.
文摘The ability to accurately predict urban traffic flows is crucial for optimising city operations.Consequently,various methods for forecasting urban traffic have been developed,focusing on analysing historical data to understand complex mobility patterns.Deep learning techniques,such as graph neural networks(GNNs),are popular for their ability to capture spatio-temporal dependencies.However,these models often become overly complex due to the large number of hyper-parameters involved.In this study,we introduce Dynamic Multi-Graph Spatial-Temporal Graph Neural Ordinary Differential Equation Networks(DMST-GNODE),a framework based on ordinary differential equations(ODEs)that autonomously discovers effective spatial-temporal graph neural network(STGNN)architectures for traffic prediction tasks.The comparative analysis of DMST-GNODE and baseline models indicates that DMST-GNODE model demonstrates superior performance across multiple datasets,consistently achieving the lowest Root Mean Square Error(RMSE)and Mean Absolute Error(MAE)values,alongside the highest accuracy.On the BKK(Bangkok)dataset,it outperformed other models with an RMSE of 3.3165 and an accuracy of 0.9367 for a 20-min interval,maintaining this trend across 40 and 60 min.Similarly,on the PeMS08 dataset,DMST-GNODE achieved the best performance with an RMSE of 19.4863 and an accuracy of 0.9377 at 20 min,demonstrating its effectiveness over longer periods.The Los_Loop dataset results further emphasise this model’s advantage,with an RMSE of 3.3422 and an accuracy of 0.7643 at 20 min,consistently maintaining superiority across all time intervals.These numerical highlights indicate that DMST-GNODE not only outperforms baseline models but also achieves higher accuracy and lower errors across different time intervals and datasets.
