Graph Neural Networks Research Articles

Automated classification of tree species using graph structure data and neural networks

Automated classification of tree species using graph structure data and neural networks

Heterogeneous Graph Neural Network with Hierarchical Attention for Group-Aware Paper Recommendation in Scientific Social Networks

In recent years, the academic groups established in Scientific Social Networks (SSNs) have not only facilitated collaboration among researchers but also enriched the relations in SSNs, providing valuable information for paper recommendation tasks. However, existing paper recommendation methods rarely consider group information and they fail to fully leverage the group information due to the heterogeneous and complex relations between researchers, papers, and groups. In this paper, a heterogeneous graph neural network with hierarchical attention, named HHA-GPR, is proposed for group-aware paper recommendation. Firstly, a heterogeneous graph is constructed based on the interactions of researchers, papers, and groups in SSNs. Secondly, a random walk-based sampling strategy is utilized to sample highly correlated heterogeneous neighbors for researchers and papers. Thirdly, a hierarchical attention network with intra-type and inter-type attention mechanisms is designed to aggregate the sampled neighbors and comprehensively model the complex relations among the heterogeneous neighbors. More specifically, an intra-type attention mechanism is introduced to aggregate the neighbors of the same type, and an inter-type attention mechanism is employed to combine the embeddings of different types to form the ultimate node embedding. Extensive experiments are conducted on the real-world CiteULike and AMiner datasets, and the experimental results demonstrate that our proposed method outperforms other benchmark methods with an average improvement of 5.3% in Precision, 5.6% in Recall, and 5.1% in Normalized Discounted Cumulative Gain (NDCG) across both datasets.

An objective quantitative diagnosis of depression using a local-to-global multimodal fusion graph neural network

An objective quantitative diagnosis of depression using a local-to-global multimodal fusion graph neural network

Graph Rewiring and Preprocessing for Graph Neural Networks Based on Effective Resistance

Graph Rewiring and Preprocessing for Graph Neural Networks Based on Effective Resistance

Self-adjusting resilient control plane for virtual software-defined optical networks

Optical networks must promptly respond to failures and efficiently handle dynamic traffic in order to fulfill their role as a critical infrastructure. Leveraging network softwarization and virtualization, virtual software-defined networks offer sufficient flexibility towards this goal by sharing the physical infrastructure among multiple tenants whose traffic must traverse the network hypervisor. In a resilient optical control plane each switch must be assigned to a primary and backup hypervisor instance through short control paths, which challenge will be addressed in this paper. First, we propose an intelligent greedy hypervisor placement heuristic which maximizes acceptance ratio for current, and preparedness for future requests. Secondly, we introduce a graph neural network model that can be seamlessly integrated with either our integer linear program or heuristic method to yield high-quality placements in significantly less time compared to our prior solutions. This enhancement renders our approach applicable to larger networks, significantly expanding its practical utility. Finally, we propose a self-adjusting hypervisor migration strategy, which continuously adapts the placement to the dynamically changing virtual network requests, thus, ensuring service continuity by avoiding frequent control plane reconfigurations. Through simulations we show that our hypervisor placement and migration strategies provide a balanced control load while they can handle a wide variety of changes.

Physics-informed line graph neural network for power flow calculation.

Power flow calculation plays a significant role in the operation and planning of modern power systems. Traditional numerical calculation methods have good interpretability but high time complexity. They are unable to cope with increasing amounts of data in power systems; therefore, many machine learning based methods have been proposed for more efficient power flow calculation. Despite the good performance of these methods in terms of computation speed, they often overlook the importance of transmission lines and do not fully consider the physical mechanisms in the power systems, thereby weakening the prediction accuracy of power flow. Given the importance of the transmission lines as well as to comprehensively consider their mutual influence, we shift our focus from bus adjacency relationships to transmission line adjacency relationships and propose a physics-informed line graph neural network framework. This framework propagates information between buses and transmission lines by introducing the concepts of the incidence matrix and the line graph matrix. Based on the mechanics of the power flow equations, we further design a loss function by integrating physical information to ensure that the output results of the model satisfy the laws of physics and have better interpretability. Experimental results on different power grid datasets and different scenarios demonstrate the accuracy of our proposed model.

MAPPING: debiasing graph neural networks for fair node classification with limited sensitive information leakage

MAPPING: debiasing graph neural networks for fair node classification with limited sensitive information leakage

Graph neural networks for strut-based architected solids

Graph neural networks for strut-based architected solids

Network community detection via neural embeddings

Recent advances in machine learning research have produced powerful neural graph embedding methods, which learn useful, low-dimensional vector representations of network data. These neural methods for graph embedding excel in graph machine learning tasks and are now widely adopted. However, how and why these methods work—particularly how network structure gets encoded in the embedding—remain largely unexplained. Here, we show that node2vec—shallow, linear neural network—encodes communities into separable clusters better than random partitioning down to the information-theoretic detectability limit for the stochastic block models. We show that this is due to the equivalence between the embedding learned by node2vec and the spectral embedding via the eigenvectors of the symmetric normalized Laplacian matrix. Numerical simulations demonstrate that node2vec is capable of learning communities on sparse graphs generated by the stochastic blockmodel, as well as on sparse degree-heterogeneous networks. Our results highlight the features of graph neural networks that enable them to separate communities in the embedding space.

A Hybrid Recognition Framework for Highly Interacting Machining Features Based on Primitive Decomposition, Learning and Reconstruction

For the highly interacting machining features, Layered Projection Decomposition Method presents inferior recognition efficiency and accuracy, due to its high-cost 3D projection and failures in determining projection faces for internal occluded faces. To address these issues, we propose a potential hybrid recognition framework. We first introduce a straightforward adjacent projection wire (APW) over UV wires, automatically restoring the full projection wires from highly interacting features. Building on APWs, an efficient hybrid boundary representation and its corresponding unambiguous primitive definitions are proposed by combing with graph-based boundary representations. Subsequently, we design an efficient primitive decomposition method by introducing primitive boundary matching to decide the initial projection faces, and introducing iterative projection boundary expansion to complete the full primitives from occluded faces. Moreover, we establish an efficient Graph Neural Network to learn the distinguishable distributions over the decomposed primitives. Specifically, an Adjacency Attention Unit is proposed to automatically perceive the influence weight of adjacent nodes, leading to more discriminative self-adaptive shape embedding for efficient primitive recognition. Finally, we summarize convenient reconstruction rules to correct the wrong predictions of feature faces with indistinguishable adjacent relationships. To evaluate the effectiveness of the proposed recognition framework, CAD models of complex aircraft structural parts are collected to present a challenging machining feature dataset. Extensive numerical experiments demonstrate that the proposed hybrid recognition framework enables significant improvements over the state-of-the-art machining feature recognition techniques.

HiMolformer: Integrating graph and sequence representations for predicting liver microsome stability with SMILES

In the initial stages of drug discovery or pre-clinical studies, understanding the metabolic stability of new molecules is crucial. Recently, research on pre-trained deep learning for molecular property prediction has been actively progressing, with various models being made open-source. However, most of these models rely on either 2D graph or 1D sequence for training, and the representation varies depending on the data format used. Consequently, combining multiple representations can broaden the scope of learning and may potentially be a manageable and most effective method to enhance performance.Therefore, we propose a novel hybrid model for predicting metabolic stability, which integrates representations from both graph-based and sequence-based models pre-trained for molecular features. This approach utilizes the combined strengths of 2D topological and 1D sequential information of molecules. HiMol, a graph-based graph neural network (GNN) model, and Molformer, a sequence-based Transformer model, were selected for integration, thus we named it HiMolformer. HiMolformer demonstrated superior performance compared to other models. We also focus on regression task for prediction with a empirical dataset from Korea Chemical Bank (KCB), comprising 3,498 molecules with mouse liver microsome (MLM) and human liver microsome (HLM) data obtained from actual metabolic reaction experiments. To the best of our knowledge, it is the first attempt to develop MLM and HLM prediction models using regression with a single SMILES input. The source code of this model is available at https://github.com/YUNSEOKWOO/HiMolformer.

DMHGNN: Double multi-view heterogeneous graph neural network framework for drug-target interaction prediction

DMHGNN: Double multi-view heterogeneous graph neural network framework for drug-target interaction prediction

Robust Knowledge Adaptation for Dynamic Graph Neural Networks

Robust Knowledge Adaptation for Dynamic Graph Neural Networks

Heterogeneous graph neural network for modeling intelligent manufacturing systems

Abstract Currently, manufacturing systems have become more and more complex, often involving multiple machines, systems and processes to produce workpieces. In order to facilitate comprehensive analysis and control of manufacturing processes, the integration of connections between machine level, system level, and process level in manufacturing process modeling is needed. However, traditional graph deep learning models are unable to take into account the heterogeneity of different machines in the system when modeling manufacturing systems. To address this problem, this paper proposes a new approach: modeling manufacturing systems using the heterogeneous graph neural network sample and aggregate algorithm based on cutting-edge Bayesian neural networks and graph deep learning. This method considers the connection between different manufacturing system levels, treats machine operations as nodes, and connects different nodes through material flow and operational similarity. The effectiveness of the method is demonstrated through its application to model an aero-engine blade production line. Extensive numerical experiments show that the proposed graph modeling method is effective in expressing the heterogeneity of different machines and multi-level manufacturing processes. By integrating machine heterogeneity into the modeling of a manufacturing system, it not only facilitates comprehensive analysis and control of the manufacturing process, but also lays the foundation for cost savings and productivity improvements in the manufacturing system.

Scaling Graph Neural Networks for Large-Scale Power Systems Analysis: Empirical Laws for Emergent Abilities

Scaling Graph Neural Networks for Large-Scale Power Systems Analysis: Empirical Laws for Emergent Abilities

Advanced graph neural network-based surrogate model for granular flows in arbitrarily shaped domains

In next-generation manufacturing, the introduction of a surrogate model is crucial for constructing a digital twin. This trend is no exception in powder industry. Various mesh-free approaches such as the discrete element method (DEM) are versatile tools for facilitating simulation-based digital twins of powder industries. Unfortunately, the DEM-based high-fidelity simulation of actual industrial processes poses a significant challenge because of the prohibitive computational costs. In this context, the surrogate model has emerged as a promising technique to solve this problem. However, surrogate modeling of powder processes within arbitrarily shaped boundaries is substantially impossible. To resolve this essential problem in the existing surrogate models, we propose a novel surrogate model, namely, a signed distance function-based graph neural network (SGN), to accelerate the simulations of granular flows. In the SGN, the wall boundary is efficiently modeled by combining the graph structure with a signed distance function. Validation studies were conducted in typical powder systems, such as a sandglass, and a rotating-paddle mixer. A systematic comparison of the simulation results between the high-fidelity and SGN simulations shows that the proposed SGN accurately simulates granular dynamics, with a remarkable increase in computational efficiency. This study significantly contributes to the next-generation manufacturing in the powder industry.

Physics-Enhanced Graph Neural Networks for Soft Sensing in Industrial Internet of Things

Physics-Enhanced Graph Neural Networks for Soft Sensing in Industrial Internet of Things

MSSTGNN: Multi-Scaled Spatio-Temporal Graph Neural Networks for short- and long-term traffic prediction

Accurate traffic prediction plays a crucial role in ensuring traffic safety and minimizing property damage. The utilization of STGNNs in traffic prediction has gained significant attention from researchers aiming to capture the intricate time-varying relationships within traffic data. While existing STGNNs commonly rely on Euclidean distance to assess the similarity between nodes, which may fall short in reflecting POI or regional functions. The traffic network exhibits static from a macro perspective, whereas undergoes dynamic changes in the micro perspective. Previous researchers incorporating self-attention to capture time-varying features for constructing dynamic graphs have faced challenges in overlooking the connections between nodes due to the Softmax polarization effect, which tends to amplify extreme value differences, and fails to accurately represent the true relationships between nodes. To solve this problem, we introduce the Multi-Scaled Spatio-Temporal Graph Neural Networks (MSSTGNN), which aims to comprehensively capture characteristics within traffic from multiscale viewpoints to construct multi-perspective graphs. We employ a trainable matrix to enhance the predefined adjacency matrices and to construct an optimal dynamic graph based on both trend and period. Additionally, a graph aggregate technique is proposed to effectively merge trend and periodic dynamic graphs. TCN is developed to model nonstationary traffic data, and we leverage the skip and residual connections to increase the model depth. A two-stage learning approach and a novel MSELoss function is designed to enhance the model's performance. The experimental results demonstrate that the MSSTGNN model outperforms the existing methods, achieving state-of-the-art performances across multiple real-world datasets.

Molecular dynamics study of the effect of composition on elastic properties of silicon oxynitride films

Understanding the mechanical properties of silicon oxynitride (a-SiON), a key insulating material, is vital for electronic device design and reliability. Though the effects of fabrication conditions on a-SiON have been studied, the underlying relationship between its atomic-scale structure and mechanical properties remains unclear. This study investigates the relationship between elasticity and atomic-scale structures in a-SiON by molecular dynamics simulations with a universal graph neural network interatomic potential. The bulk modulus increases from 49 to 150 GPa with higher N content. N atoms form N2 molecules under O-rich conditions, hindering bulk modulus increase, and form an Si3N4-like network under O-poor conditions, enhancing bulk modulus. Formation energy calculations indicate N2 formation is preferable under O-rich conditions. Meanwhile, under O-poor conditions, Si–N bond formation is preferable, which reinforces a-SiON by increasing bond density. The findings suggest realizing O-poor conditions is crucial for highly elastic insulating films.

Wireless Network Security Prediction Using Machine Learning and Graph Neural Networks

Abstract: We study the application of graph neural networks and machine learning in the prediction of risks to wireless networks. Advanced predictive methods are necessary, since mere security measures cannot thwart the rising complexity of cyber-attacks. In the present work, we will use a hybrid model that combines several machine learning methods to reduce false negatives and positives and improve accuracy in the prediction of risks to wireless networks. Models are trained and validated with large amounts of data that include the performance indicators for accuracy, precision, recall, F1-score, and AUC-ROC. From this, it is also seen that the hybrid model performs much better than the standard models in real-time threat detection. The impact of the size of the dataset on the performance of the model was also studied, and it has come out that larger datasets improve predictive powers significantly. The result demonstrated that the latest advancements in machine learning techniques can lead to important improvements in the security of wireless networks.