About this Research Topic
Graphs are a ubiquitous representation and powerful data structure that can represent an array of real-world phenomena such as social networks, biological networks, transportation systems, recommendation systems, and more. For example, in other domains such as communication networks, graphs have been used, to represent network components and their relations, routing networks and their attributes, traffic and its variations; to name a few.
Graph Neural Networks (GNNs) are a recent family of Neural Network models specifically designed to harness the inherent structure and dependencies present in graph-structured data, revolutionizing the way we analyze, model, and make predictions in complex networked structures. Unlike traditional machine learning models, which often struggle with capturing such relationships, GNNs excel at encoding the intricate interconnections within data points. They offer significant advantages over conventional neural architectures, demonstrating ground-breaking applications, where data is generated from non-Euclidean domains, and represented as graphs with complex relationships and interdependencies between objects.
GNNs owe their success to their innovative architecture which exploits the topological structure of graphs and enables the aggregation of information from neighboring nodes iteratively, allowing them to capture multi-hop dependencies and learn the feature vector of all nodes. Convolutional and recurrent operations, adapted to work on graph structures, lie at the heart of GNNs. Researchers have continuously pushed the boundaries of GNN design, leading to a rich ecosystem of variants, including Graph Convolutional Networks (GCNs), GraphSAGE , Graph Attention Networks (GATs), and many others. The use of such variants and learning methods, such as network embeddings and representation learning, have led to unprecedented progress in solving many challenges facing real-world applications, such as recommender systems, anomaly detection, smart environments, traffic forecasting, disease control and prevention, medical diagnosis, and drug discovery.
This Research Topic aims to cover applications where GNNs have proven to be effective. We invite authors from academia and industry to contribute their original research articles, surveys, and high-quality review papers, that demonstrate the effectiveness of GNNs in solving real-world problems while showcasing the latest developments and novel applications. We encourage submissions that address, but are not limited to, the following areas:
• Graph Embeddings
• Deep Learning on Graphs (Graph Convolutions, Graph Attention Networks, Graph Autoencoders, Graph Spatial-Temporal Networks)
• Network Representation Learning
• Learning on dynamic, temporal, and/or complex graphs
• Knowledge modeling/representation in/for graph learning
• Graph datasets and benchmarks
• Novel models and algorithms for graphs
• Node/Graph Classification/Prediction
• Graph Clustering/Visualisation
Applications:
• Wired/Wireless Communication Networks
• IoT
• Bioinformatics
• Physics
• Natural Language Processing
• Computer Vision
• Knowledge Graphs
• Recommendation Systems
• Other areas (Mobility/Transportation, Geographical, Financial, Robotics/Cyber-physical)
Graph Neural Networks (GNNs) are a recent family of Neural Network models specifically designed to harness the inherent structure and dependencies present in graph-structured data, revolutionizing the way we analyze, model, and make predictions in complex networked structures. Unlike traditional machine learning models, which often struggle with capturing such relationships, GNNs excel at encoding the intricate interconnections within data points. They offer significant advantages over conventional neural architectures, demonstrating ground-breaking applications, where data is generated from non-Euclidean domains, and represented as graphs with complex relationships and interdependencies between objects.
GNNs owe their success to their innovative architecture which exploits the topological structure of graphs and enables the aggregation of information from neighboring nodes iteratively, allowing them to capture multi-hop dependencies and learn the feature vector of all nodes. Convolutional and recurrent operations, adapted to work on graph structures, lie at the heart of GNNs. Researchers have continuously pushed the boundaries of GNN design, leading to a rich ecosystem of variants, including Graph Convolutional Networks (GCNs), GraphSAGE , Graph Attention Networks (GATs), and many others. The use of such variants and learning methods, such as network embeddings and representation learning, have led to unprecedented progress in solving many challenges facing real-world applications, such as recommender systems, anomaly detection, smart environments, traffic forecasting, disease control and prevention, medical diagnosis, and drug discovery.
This Research Topic aims to cover applications where GNNs have proven to be effective. We invite authors from academia and industry to contribute their original research articles, surveys, and high-quality review papers, that demonstrate the effectiveness of GNNs in solving real-world problems while showcasing the latest developments and novel applications. We encourage submissions that address, but are not limited to, the following areas:
• Graph Embeddings
• Deep Learning on Graphs (Graph Convolutions, Graph Attention Networks, Graph Autoencoders, Graph Spatial-Temporal Networks)
• Network Representation Learning
• Learning on dynamic, temporal, and/or complex graphs
• Knowledge modeling/representation in/for graph learning
• Graph datasets and benchmarks
• Novel models and algorithms for graphs
• Node/Graph Classification/Prediction
• Graph Clustering/Visualisation
Applications:
• Wired/Wireless Communication Networks
• IoT
• Bioinformatics
• Physics
• Natural Language Processing
• Computer Vision
• Knowledge Graphs
• Recommendation Systems
• Other areas (Mobility/Transportation, Geographical, Financial, Robotics/Cyber-physical)
Keywords: Graphs, Complex Networks, Graph Learning, Applications, Communication Networks, Social Networks, Biological Networks, IoT, Physics Simulation, Recommendation Systems, Smart Environments, Drug Discovery, graph neural networks
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
