A Comprehensive Survey on Graph Neural Networks
By Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, Philip S. Yu. 2019
11,551 citations recorded by the source corpus.
Deep learning has revolutionized many machine learning tasks in recent years, ranging from image classification and video processing to speech recognition and natural language understanding. The data in these tasks are typically represented in the Euclidean space. However, there is an increasing number of applications, where data are generated from non-Euclidean domains and are represented as graphs with complex relationships and interdependency between objects. The complexity of graph data has imposed significant challenges on the existing machine learning algorithms. Recently, many studies on extending deep learning approaches for graph data have emerged. In this article, we provide a comprehensive overview of graph neural networks (GNNs) in data mining and machine learning fields. We propose a new taxonomy to divide the state-of-the-art GNNs into four categories, namely, recurrent GNNs, convolutional GNNs, graph autoencoders, and spatial–temporal GNNs. We further discuss the applications of GNNs across various domains and summarize the open-source codes, benchmark data sets, and model evaluation of GNNs. Finally, we propose potential research directions in this rapidly growing field.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Graph Learning, Evaluation & Benchmarks
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Yoshua Bengio
- Percy Liang
- Mohit Bansal
- Jianfeng Gao
- Quoc V. Le
- Luke S. Zettlemoyer
- Dacheng Tao
- Sergey Levine
In Computer Vision and Pattern Recognition
- Dacheng Tao
- Luc Van Gool
- Xiaogang Wang
- Trevor Darrell
- Ming-Hsuan Yang
- Yi Yang
- Jitendra Malik
- Andrea Vedaldi
In Graph Learning
- Jure Leskovec
- William L. Hamilton
- Michael M. Bronstein
- Marinka Zitnik
- Xavier Bresson
- Bryan Perozzi
- Shirui Pan
- Stephan Günnemann
In Evaluation & Benchmarks
Citations in the corpus
- On the Binding Problem in Artificial Neural Networks (2020, background): “Graph Neural Networks (GNNs; Scarselli et al., 2009; Pollack, 1990) are a promising approach for composition that incorporates the desired structure for relational responding at an architectural level (see Wu et al. 2020 for an overview).”
- Counterfactual Multi-Agent Reinforcement Learning with Graph Convolution Communication (2020, neutral): “Graph Convolution : Graph convolution extends the application of convolution neural networks from the euclidean domain to the non-euclidean domain such as graphs [24].”
- Unsupervised Inductive Graph-Level Representation Learning via Graph-Graph Proximity (2019, background): “Categorized as “graph autoencoders” [Wu et al., 2019], they also belong to the family of graph neural network methods.”
- Graph Warp Module: an Auxiliary Module for Boosting the Power of Graph Neural Networks (2019, background): “Graph Neural Network is a family of networks specializing in processing the features of graph-structured data (see [34, 30] for comprehensive reviews), and there has been a surge of interest in its application to the analysis of social networks and chemical structure.”
- Lifelong Graph Learning (2020, neutral): “Graph neural networks (GNN) have received increasing attention and proved useful for many tasks with graph-structured data, such as citation, social, and protein networks [51].”
- Simplifying Graph Convolutional Networks (2019, background): “We refer to Zhou et al. (2018); Battaglia et al. (2018); Wu et al. (2019) for a more comprehensive review.”
- Learning to Fix Build Errors with Graph2Diff Neural Networks (2019, background): “There has been much recent work on graph neural networks (GNNs) [41] but less work on using them to generate sequential outputs.”
- Geometric deep learning for computational mechanics Part I: Anisotropic Hyperelasticity (2020, background): “An extensive summary of different graph neural network architectures currently developed can be found in (Wu et al., 2019).”
- Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition (2020, background): “Among different GNN variants, the Graph Convolutional Network (GCN) [17] was first introduced as a first-order approximation for localized spectral convolutions, but its simplicity as a mean neighborhood aggregator [49, 46] has quickly led many subsequent spatial GNN architectures [49, 1, 45, 7] and”
- Graph Convolutional Networks with EigenPooling (2019, background): “Comprehensive surveys on graph neural networks can be found in [2, 46, 50, 51].”
- The Deep Learning Compiler: A Comprehensive Survey (2020, background): “Graph neural network (GNN) - GNN has been a popular research direction in the field of deep learning in recent years [22, 105, 107, 110].”
- Learning to Dispatch for Job Shop Scheduling via Deep Reinforcement Learning (2020, background): “A natural and straightforward strategy for this is to replace each undirected arc in G(t) by two directed ones connecting the same nodes with opposite directions, resulting in a fully directed graph denoted as GD(t) [39, 40].”
- HyGCN: A GCN Accelerator with Hybrid Architecture (2020, background): “GCN is one of the most successful convolutional networks for graph learning [2, 12], which bridges the gap between spectral-based convolutions and spatial-based convolutions.”
- Answering Any-hop Open-domain Questions with Iterative Document Reranking (2020, background): “Encouraged by the success of convolutional neural networks (CNNs) and recurrent neural networks (RNNs), graph neural networks (GNNs) are proposed to generalize both of them and organize the connections of neurons according to the structure of the input data [2, 38].”
- A Comprehensive Survey on Graph Anomaly Detection With Deep Learning (2021, background): “The more recent studies, such as deep graph representation learning and graph neural networks (GNNs), further enrich the capability of deep learning for graph data mining [58], [59], [60], [61], [62].”
- A Survey of Knowledge Graph Reasoning on Graph Types: Static, Dynamic, and Multi-Modal (2022, background): “3) GNN Model: Graph neural networks (GNN) [72], [73] are widely used for graph tasks, due to their promising capacity for structural information mining.”
- Knowledge Graphs (2020, positive): “Most such architectures are based on neural networks [145] given that a neural network is already a directed weighted graph, where nodes serve as artificial neurons, and edges serve as weighted connections (axons).”
- Deep Neural Networks and Tabular Data: A Survey (2021, methodology): “Thus, the proposed framework exploits the topology information from the decision trees using graph neural networks [103].”
- Graph Neural Networks for Scalable Radio Resource Management: Architecture Design and Theoretical Analysis (2020, background): “This motivates researchers to develop GNNs [29], which effectively exploit the graph structure.”