Neighbor Nodes Research Articles

Neighborhood contrastive learning-based graph neural network for bug triaging

Researchers have been developing automatic bug triaging techniques by leveraging bug information sourced from bug tracking systems. Recent studies have modeled the bug-developer relationship as a graph, introducing graph neural networks for bug triaging. Despite achieving positive outcomes, these methods overlook issues related to data sparsity and fail to fully exploit implicit relationships within the graph. In addressing these challenges, we present the Neighborhood Contrastive Learning-based Graph Neural Network Bug Triaging framework, abbreviated as NCGBT. Our approach models the relationship between bugs and developers as a bipartite graph. We utilize a pre-trained language model to acquire the initial representation of bug nodes. Employing a basic graph neural network framework, we learn the representation of all nodes and leverage these representations to predict developers for a given bug. Our proposed strategy involves neighborhood contrastive learning applied to the basic graph neural network approach. We take into account the neighbors of nodes from both structural and semantic perspectives, utilizing them as contrastive objects. Extensive experiments conducted on three public datasets demonstrate the effectiveness of the NCGBT framework.

GNNCL: A Graph Neural Network Recommendation Model Based on Contrastive Learning

In the field of recommendation algorithms, the representation learning for users and items has evolved from using single IDs or historical interactions to utilizing higher-order neighbors. This can be achieved by modeling the user–item interaction graph to capture user preferences for items. Despite the promising results achieved by these algorithms, they still suffer from the issue of data sparsity. In order to mitigate the impact of data sparsity, contrastive learning has been adopted in graph collaborative filtering to enhance performance. However, current recommendation algorithms using contrastive learning yield uneven representations after data augmentation and do not consider the potential relationships among users (or items). To address these challenges, we propose a graph neural network-based recommendation model that integrates contrastive learning (GNNCL). This model combines data augmentation with added noise and the exploration of semantic neighbors for nodes. For the structural neighbors on the interaction graph, we introduce a novel and straightforward contrastive learning approach, abandoning previous graph augmentation methods, and introducing uniform noise into the embedding space to create contrastive views. To unearth potential semantic neighbor relationships in the semantic space, we assume that users with similar representations possess semantic neighbor relationships and merge these semantic neighbors into the prototype contrastive learning. We utilize a clustering algorithm to obtain prototypes for users and items and employ the EM algorithm for prototype contrastive learning. Experimental results validate the effectiveness of our approach. Particularly, on the Yelp2018 and Amazon-book datasets, our method exhibits significant performance improvements compared to basic graph collaborative filtering models.

Curvature-enhanced graph convolutional network for biomolecular interaction prediction

Geometric deep learning has demonstrated a great potential in non-Euclidean data analysis. The incorporation of geometric insights into learning architecture is vital to its success. Here we propose a curvature-enhanced graph convolutional network (CGCN) for biomolecular interaction prediction. Our CGCN employs Ollivier-Ricci curvature (ORC) to characterize network local geometric properties and enhance the learning capability of GCNs. More specifically, ORCs are evaluated based on the local topology from node neighborhoods, and further incorporated into the weight function for the feature aggregation in message-passing procedure. Our CGCN model is extensively validated on fourteen real-world bimolecular interaction networks and analyzed in details using a series of well-designed simulated data. It has been found that our CGCN can achieve the state-of-the-art results. It outperforms all existing models, as far as we know, in thirteen out of the fourteen real-world datasets and ranks as the second in the rest one. The results from the simulated data show that our CGCN model is superior to the traditional GCN models regardless of the positive-to-negative-curvature ratios, network densities, and network sizes (when larger than 500).

A secure edge computing model using machine learning and IDS to detect and isolate intruders

The article presents a secure edge computing model that utilizes machine learning for intrusion detection and isolation. It addresses the security challenges arising from the rapid expansion of IoT and edge computing. The proposed Intrusion Detection System (IDS) combines Linear Discriminant Analysis (LDA) and Logistic Regression (LR) to swiftly and accurately identify intrusions without alerting neighboring devices. The model outperforms existing solutions with an accuracy of 96.56%, precision of 95.78%, and quick training time (0.04 s). It is effective against various types of attacks, enhancing the security of edge networks for IoT applications.•The methodology employs a hybrid model that combines LDA and LR for intrusion detection.•Machine learning techniques are used to analyze and identify intrusive activities during data acquisition by edge nodes.•The methodology includes a mechanism to isolate suspected devices and data without notifying neighboring edge nodes to prevent intruders from gaining control over the edge network.

Fuzzy adaptive containment control for fractional‐order heterogeneous nonlinear multi‐agent systems with mixed time‐varying delays

SummaryThis study investigates the issue of containment control (CC) of multi‐agent systems (MASs) that are heterogeneous, nonlinear, and fractional‐order while taking practical scenarios into account, such as uncertainties, unknown nonlinearities, time‐varying state delays, and distributed time‐varying delays. A fully distributed adaptive observer is created for each follower in accordance with the communication topology information among agents. This observer is designed to estimate the convex combination information of the leaders. The approach utilizes interval type‐2 fuzzy set theory and adaptive methods to design a distributed containment control strategy that only uses the self‐information of the followers and neighbor nodes' information. Sufficient conditions for implementing containment control are given. A new Lyapunov‐Krasovskii functional (LKF) is constructed, and the stability of the system with feedback loop is proven using fractional‐order theory and inequality methods. Lastly, a simulation example is presented to illustrate the efficacy of the proposed approach.

Conversational recommender based on graph sparsification and multi-hop attention

Conversational recommender systems provide users with item recommendations via interactive dialogues. Existing methods using graph neural networks have been proven to be an adequate representation of the learning framework for knowledge graphs. However, the knowledge graph involved in the dialogue context is vast and noisy, especially the noise graph nodes, which restrict the primary node’s aggregation to neighbor nodes. In addition, although the recurrent neural network can encode the local structure of word sequences in a dialogue context, it may still be challenging to remember long-term dependencies. To tackle these problems, we propose a sparse multi-hop conversational recommender model named SMCR, which accurately identifies important edges through matching items, thus reducing the computational complexity of sparse graphs. Specifically, we design a multi-hop attention network to encode dialogue context, which can quickly encode the long dialogue sequences to capture the long-term dependencies. Furthermore, we utilize a variational auto-encoder to learn topic information for capturing syntactic dependencies. Extensive experiments on the travel dialogue dataset show significant improvements in our proposed model over the state-of-the-art methods in evaluating recommendation and dialogue generation.

Learning Embedding for Signed Network in Social Media With Global Information

Signed networks contain nodes connected by positive and negative signed links. Signed network representation learning concentrates on learning the low-dimensional representations of these nodes, which facilitates downstream tasks such as link prediction using general data mining framework. However, most signed network embedding (SiNE) approaches neglect global information in the networks, limiting the capacity to learn genuine signed graph topology. To overcome this limitation, a deep graph neural network (GNN) framework SiG to learn SiNE with global information is proposed. To be more explicit, a hierarchical pooling mechanism is developed to encode the high-level features of the networks. Moreover, a graph convolution layer is introduced to aggregate both positive and negative information from neighbor nodes, and the concatenation of two parts generates the final embedding of nodes. Extensive experiments on four large real-world signed network datasets demonstrate the effectiveness and excellence of the proposed method.

A novel self-adaptive SIS model based on the mutual interaction between a graph and its line graph.

We propose a new paradigm to design a network-based self-adaptive epidemic model that relies on the interplay between the network and its line graph. We implement this proposal on a susceptible-infected-susceptible model in which both nodes and edges are considered susceptible and their respective probabilities of being infected result in a real-time re-modulation of the weights of both the graph and its line graph. The new model can be considered as an appropriate perturbation of the standard susceptible-infected-susceptible model, and the coupling between the graph and its line graph is interpreted as a reinforcement factor that fosters diffusion through a continuous adjustment of the parameters involved. We study the existence and stability conditions of the endemic and disease-free states for general network topologies. Moreover, we introduce, through the asymptotic values in the endemic steady states, a new type of eigenvector centrality where the score of a node depends on both the neighboring nodes and the edges connected to it. We also investigate the properties of this new model on some specific synthetic graphs, such as cycle, regular, and star graphs. Finally, we perform a series of numerical simulations and prove their effectiveness in capturing some empirical evidence on behavioral adoption mechanisms.

New Method of Edge Computing-Based Data Adaptive Return in Internet of Vehicles

Edge computing technology can be used to the Internet of Vehicles (IOV) to solve the mobile characteristics of vehicles and the limited communication range between roadside units and vehicles. New method of edge computing based data adaptive return in IOV is proposed in this paper. The transmission strategy can be determined by adaptive estimating the vehicle movement, the amount of data returned, the maximum transmission delay and effective life of link. And the factors such as speed, direction and position of the vehicles are comprehensively considered and these factors can be measured by the stability effect value when adaptive designing the auxiliary transmission strategy. At the same time, greedy selection method (GSM) is used when constructing the data return link, and the neighbor node as the relay node with the maximum stability and efficiency value is chosen. Our experimental results show our method in terms of performance on transmission delay & packet delivery rate is better than other ones.

A Heterogeneous Graph Neural Network With Attribute Enhancement and Structure-Aware Attention

Heterogeneous information network (HIN) has been applied in a wide variety of graph analysis tasks. At present, it is a trend of heterogeneous graph neural networks (HGNNs) to cast the meta-paths aside, since it solves the problem of structural information loss caused by artificially designed meta-paths. However, existing meta-path-free HGNNs fail to take into account that most node types in many HINs have no attributes, and they cannot make full use of sparse node attributes when applied to HINs with missing attributes. Furthermore, their computation of attention coefficients explores the correlations of node attributes while almost ignoring structural ones, which may limit the expression ability of the model and cause overfitting in model training. To alleviate these issues, we propose an HGNN with attribute enhancement and structure-aware attention (HGNN-AESA). First, we design an attribute enhancement module (AEM) to connect more useful attributed nodes to the target nodes. Specifically, AEM introduces a random walk with restart (RWR) strategy to obtain structural relevance scores of each node within its specific subgraph. The structural relevance scores are used to capture potentially influential attributed nodes in high-order neighborhood for each target node. Second, we propose heterogeneous structure-aware attention layers (HSALs) to learn node representations. HSALs follow a hierarchical attention framework, including node-level and type-level attention. The node-level attention aggregates feature (attribute) embeddings of same-type neighbors, and the relevant attention coefficients depend on the combination of node attributes and heterogeneous structural interventions. The type-level attention fuses all type-specific vector representations and generates the ultimate node embedding. Finally, extensive experiments on three different real-world HIN datasets demonstrate that our model outperforms state-of-the-art methods.

Q-DDCA: Decentralized Dynamic Congestion Avoid Routing in Large-Scale Quantum Networks

The quantum network that allows users to communicate in a quantum way will be available in the foreseeable future. The network capable of distributing Bell state entangled pairs faces many challenges due to entanglement decoherence and limited network performance, especially when the network scale is enormous. Many entanglement distribution protocols have been proposed so far, and most of them are in a centralized and synchronized manner, which may be infeasible in large-scale networks. As such, in this paper, we propose a full spontaneous version of quantum networks in which the quantum nodes autonomously manage multiple entanglement distribution requests. However, one major issue is that quantum nodes have little knowledge about the network, especially the congestion (e.g., some nodes may have no usable quantum memories). We present a routing algorithm to adaptive evaluate the congestion on the neighbor nodes to avoid potential congestion. We use SimQN, the new network layer simulation platform built by our research team, to evaluate our proposed design. The result demonstrates that it can adapt to changes in network resources and reduce the drop rate that eventually leads to a higher entanglement distribution rate but remains fair for multiple requests to use the network resources fairly and achieve a more balanced throughput.

HGSMDA: miRNA-Disease Association Prediction Based on HyperGCN and Sørensen-Dice Loss.

Biological research has demonstrated the significance of identifying miRNA-disease associations in the context of disease prevention, diagnosis, and treatment. However, the utilization of experimental approaches involving biological subjects to infer these associations is both costly and inefficient. Consequently, there is a pressing need to devise novel approaches that offer enhanced accuracy and effectiveness. Presently, the predominant methods employed for predicting disease associations rely on Graph Convolutional Network (GCN) techniques. However, the Graph Convolutional Network algorithm, which is locally aggregated, solely incorporates information from the immediate neighboring nodes of a given node at each layer. Consequently, GCN cannot simultaneously aggregate information from multiple nodes. This constraint significantly impacts the predictive efficacy of the model. To tackle this problem, we propose a novel approach, based on HyperGCN and Sørensen-Dice loss (HGSMDA), for predicting associations between miRNAs and diseases. In the initial phase, we developed multiple networks to represent the similarity between miRNAs and diseases and employed GCNs to extract information from diverse perspectives. Subsequently, we draw into HyperGCN to construct a miRNA-disease heteromorphic hypergraph using hypernodes and train GCN on the graph to aggregate information. Finally, we utilized the Sørensen-Dice loss function to evaluate the degree of similarity between the predicted outcomes and the ground truth values, thereby enabling the prediction of associations between miRNAs and diseases. In order to assess the soundness of our methodology, an extensive series of experiments was conducted employing the Human MicroRNA Disease Database (HMDD v3.2) as the dataset. The experimental outcomes unequivocally indicate that HGSMDA exhibits remarkable efficacy when compared to alternative methodologies. Furthermore, the predictive capacity of HGSMDA was corroborated through a case study focused on colon cancer. These findings strongly imply that HGSMDA represents a dependable and valid framework, thereby offering a novel avenue for investigating the intricate association between miRNAs and diseases.

Meet in the air: Distributed neighbor discovery in 3D networks with directional transceivers

We formulate and analyze a generic neighbor discovery problem in three-dimensional wireless networks with directional transceivers, where any pair of neighbor nodes in 3D Euclidean space need to steer their transceivers towards each other simultaneously to discover each other. This problem, termed as 3D directional neighbor discovery, arises in a variety of emerging networked systems such as flying ad hoc networks composed of drones and distributed space systems composed of spacecrafts. Compared to the omni-directional 2D neighbor discovery problem extensively investigated in the literature, the 3D directional neighbor discovery problem is intuitively more challenging. Motivated by this observation, we establish an algorithmic framework on the 3D directional neighbor discovery problem. We first mathematically formulate the problem and derive the worst-case discovery delay of any neighbor discovery algorithm. Guided by the performance limit, we then design distributed neighbor discovery algorithms achieving bounded and minimal worst-case discovery delay. We further demonstrate how our algorithmic framework can be generalized to solve generic multi-dimensional rendezvous problems. Extensive numerical analysis is then presented to empirically evaluate of our neighbor discovery algorithms.

Virtual Coordinate System Based on a Circulant Topology for Routing in Networks-On-Chip

In this work, the circulant topology as an alternative to 2D mesh in networks-on-chip is considered. A virtual coordinate system for numbering nodes in the circulant topology is proposed, and the principle of greedy promotion is formulated. The rules for constructing the shortest routes between the two nodes based on coordinates are formulated. A technique for calculating optimal network configurations is described. Dense states of the network when all neighborhoods of the central node are filled with nodes and the network has the smallest diameter are defined. It is shown that with an equal number of nodes, the diameter of the circulant is two times smaller than the diameter of the 2D mesh. This is due to the large number of symmetries for the circulant, which leave the set of nodes unchanged. A comparison of communication stability in both topologies in the conditions of failure of network nodes is made, the network behavior under load and failures is modeled, and the advantages of the circulant topology are presented.

Identification of Critical Links Based on Electrical Betweenness and Neighborhood Similarity in Cyber-Physical Power Systems.

Identifying critical links is of great importance for ensuring the safety of the cyber-physical power system. Traditional electrical betweenness only considers power flow distribution on the link itself, while ignoring the local influence of neighborhood links and the coupled reaction of information flow on energy flow. An identification method based on electrical betweenness centrality and neighborhood similarity is proposed to consider the internal power flow dynamic influence existing in multi-neighborhood nodes and the topological structure interdependence between power nodes and communication nodes. Firstly, for the power network, the electrical topological overlap is proposed to quantify the vulnerability of the links. This approach comprehensively considers the local contribution of neighborhood nodes, power transmission characteristics, generator capacity, and load. Secondly, in communication networks, effective distance closeness centrality is defined to evaluate the importance of communication links, simultaneously taking into account factors such as the information equipment function and spatial relationships. Next, under the influence of coupled factors, a comprehensive model is constructed based on the dependency relationships between information flow and energy flow to more accurately assess the critical links in the power network. Finally, the simulation results show the effectiveness of the proposed method under dynamic and static attacks.

Identifying the critical nodes in multi-modal transportation network with a traffic demand-based computational method

In the last decade, there has been an abundance of research on identifying critical nodes in single-modal transportation network, but less for multi-modal transportation network (MTN). In addition, none of them consider the effect of interactions between nodes in a transportation corridor on node importance. For this purpose, a modified weighted k-shell (MWKS) model is proposed for identifying the critical nodes of the MTN, which combines the network topology and traffic demand. It is modified by the differential contribution of the multi-order neighbor nodes embodied by the passenger volume and the transportation corridor idea. Then the effectiveness of MWKS is verified with an example of a real transportation system using susceptible-infected (SI) model and cascading failure model. The analysis proves that MWKS is suitable for identifying critical nodes in the MTN. Furthermore, the proposed contribution attenuation coefficient based on the idea of the transportation corridor outperforms the topology-based and constant. It demonstrates the necessity of capturing the difference in the contribution of each neighbor node based on traffic demand when calculating the node importance. The results can provide a research basis for traffic planning to improve network reliability.

Enabling Graph Neural Networks for Semi-Supervised Risk Prediction in Online Credit Loan Services

Graph neural networks (GNNs) are playing exciting roles in the application scenarios where features are hidden in information associations. Fraud prediction of online credit loan services (OCLSs) is such a typical scenario. But it has another rather critical challenge, i.e., the scarcity of data labels. Fortunately, GNNs can also cope with this problem due to their good ability of semi-supervised learning by mining structure and feature information within graphs. Nevertheless, the gain of internal information is often too limited to help GNNs handle the extreme deficiency of labels with high performance beyond the basic requirement of fraud prediction in OCLSs. Therefore, adding labels from the experts, such as manually adding labels through rules, has become a logical practice. However, the existing rule engines for OCLSs have the confliction problem among continuously accumulated rules. To address this issue, we propose a Snorkel-based Semi-Supervised GNN (S3GNN). Under S3GNN, we specially design an upgraded version of the rule engines, called Graph-Oriented Snorkel (GOS), a graph-specific extension of Snorkel, a widely used weakly supervised learning framework, to design rules by subject matter experts (SMEs) and resolve confliction. In particular, in the graph of an anti-fraud scenario, each node pair may have multiple different types of edges, so we propose the Multiple Edge-Types Based Attention mechanism. In general, for the heterogeneous information and multiple relations in the graph, we first obtain the embedding of applicant nodes by aggregating the representation of attribute nodes, and then use the attention mechanism to aggregate neighbor nodes on multiple meta-paths to get ultimate applicant node embedding. We conduct experiments over the real-life data of a large financial platform. The results demonstrate that S3GNN can outperform the state-of-the-art methods, including the method of pilot platform.

Influence node analysis based on neighborhood influence vote rank method in social network

Social Networks are used for various purposes like advertising, product launches, sentiment analysis, opinion mining, and event detection etc. Terrorist targets social network users to spread the terrorism. Influence analysis is used in social networks to find the influence of users and the impact of the messages, mainly for advertising. In this research, the Neighborhood Influence – Vote Rank (NI-VR) method is proposed to analyze the terrorism and social network datasets temporally to find the influence node in Social networks. The Global Terrorism Dataset (GTD) was used to analyze the terrorism activity and temporal analysis on Social Network data to find the influence node. The Neighborhood node influence is measured and considered in the Social Network data to effectively find the influence node. The nodes’ vote score and vote ability were measured to rank the nodes based on influence. The neighborhood influence is measured to update the vote score and vote ability based on influence value. The neighborhood influence method is applied to rank the node has the advantage of analyzing the probability of affected nodes and recover nodes that help to effectively find the influence nodes. The outcomes illustrate that the proposed NI-VR achieved a maximum spread influence of 843 and the existing Greedy method has a higher spread influence of 840 in influence node analysis.

Energy efficiency of cache collaboration in core content distribution networks

With escalating demands for high-definition video, cache collaboration allows neighbor nodes to share locally stored content in order to reduce download traffic. High energy consumption associated with content delivery remains a concern for Content Distribution Networks (CDNs). Therefore, this paper proposes cluster-based collaborative caching in a core network employing IP over WDM. The aim is to allow sets of core caches to fully share content while minimizing power. A Mixed Integer Linear Programming (MILP) model is used to form energy-efficient cache clusters. The energy consumption of the network is evaluated under different cluster sizes to find the optimum size that minimizes energy. To evaluate the influence of content popularity distribution, a heavy-tailed Zipf distribution and an Equal popularity distribution are evaluated. In addition, the work investigates the influence of downlink traffic behavior and power consumption parameters on optimum cluster sizes. Attained results reveal that maximum savings in energy consumption introduced by cluster-based collaborative caching are up to 34.3% and 21.8% under the Zipf and Equal distribution, respectively. Cache collaboration is not recommended when all core nodes contain fully replicated content servers. Results also show that power consumption parameters do not influence cluster formation. It is recommended keeping cache collaboration in the core network simple, so as to reduce intra-cluster communication.

Feature aggregation-based multi-relational knowledge reasoning for COPD intelligent diagnosis

The increasing prevalence of artificial intelligence-based knowledge reasoning has contributed to more accurate and efficient auxiliary diagnoses. However, a majority of the disease prediction methods concentrate on the symptoms themselves while discarding the inherent properties of symptoms and the relationships underlying them. This paper proposes a feature aggregation-based intelligent diagnosis model employing a Heterogeneous Graph Convolutional Network (GCN), termed HeteroGCN. It focuses on symptoms’ inherent properties and multiple hidden relationships among symptoms and properties. By aggregating features of nodes, it realizes effective and accurate symptom-based knowledge reasoning for disease-type prediction. The diagnosis-related information from the Electronic Medical Record (EMR) has been extracted and standardized by taking chronic obstructive pulmonary disease (COPD) as an instance. Then the presented model extracts the symptoms and their properties as nodes and the relationships underlying the nodes as edges to construct a heterogeneous graph. The adjacency matrix and feature matrix have been fused and taken as the input of this model, and then the node representations (embeddings) are generated by aggregating neighbor nodes’ information. Finally, specific disease types (syndromes) will be predicted by the generated symptom node embeddings. The results of the model comparison and parameter sensitivity test demonstrate that the presented HeteroGCN model performs best on disease-type prediction. This paper provides a novel feature aggregation-based multi-relational knowledge reasoning approach for disease type (syndrome) prediction, which holds great significance in improving disease diagnosis.