Adaptive Graph Convolutional Neural Networks

By Ruoyu Li, Sheng Wang, Feiyun Zhu, Junzhou Huang. AAAI Conference on Artificial Intelligence, 2018

867 citations recorded by the source corpus.

Graph Convolutional Neural Networks (Graph CNNs) are generalizations of classical CNNs to handle graph data such as molecular data, point could and social networks. Current filters in graph CNNs are built for fixed and shared graph structure. However, for most real data, the graph structures varies in both size and connectivity. The paper proposes a generalized and flexible graph CNN taking data of arbitrary graph structure as input. In that way a task-driven adaptive graph is learned for each graph data while training. To efficiently learn the graph, a distance metric learning is proposed. Extensive experiments on nine graph-structured datasets have demonstrated the superior performance improvement on both convergence speed and predictive accuracy.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Neurology

View Paper: https://www.semanticscholar.org/paper/51a2bc2e8fb8ed47a085df33dd965e57335080a0

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

In Computer Vision and Pattern Recognition

In Neurology

Citations in the corpus

An Introductory Review of Deep Learning for Prediction Models With Big Data (2020, methodology): “Such models have been used in natural language processing, recommender systems, genomics and chemistry (Li et al., 2018; Yao et al., 2019).”
Relation-Shape Convolutional Neural Network for Point Cloud Analysis (2019, background): “Graph convolution network is applied on a local graph created by neighboring points [38, 37, 20].”
Skeleton-Based Action Recognition With Multi-Stream Adaptive Graph Convolutional Networks (2019, background): “[39] assume that the optimal graph topology is a small shifting from the original graph topology and propose to learn a residual graph given the current data.”
Combining Graph Neural Networks With Expert Knowledge for Smart Contract Vulnerability Detection (2021, background): “[41] proposes a graph CNN which can take data of arbitrary graph structure as input.”
Self-Supervised Learning: Generative or Contrastive (2020, methodology): “In general, existing network embedding approaches are broadly categorized as (1) factorization-based approaches such as NetMF [104], [105], GraRep [18], HOPE [96], (2) shallow embedding approaches such as DeepWalk [100], LINE [128], HARP [21], and (3) neural network approaches [19], [81].”
Semi-Supervised Learning With Graph Learning-Convolutional Networks (2019, methodology): “[19] propose optimal graph CNNs, in which the graph is learned adaptively by using a distance metric learning.”
Graph Neural Networks: A Review of Methods and Applications (2018, background): “However, there may have implicit relations between different nodes and the Adaptive Graph Convolution Network (AGCN) is proposed to learn the underlying relations [49].”
Enhancing Graph Neural Network-based Fraud Detectors against Camouflaged Fraudsters (2020, methodology): “y GNN variants belong to this type. The proposed CARE-GNN is a spatial-based GNN as well. To enhance the GNN performance on graphs with noisy nodes. One approach is the graph structure learning (GSL) [6, 9, 20]. Those workslearnnewgraphstructuresfromoriginalgraphs,whichcould 1 2 3 (a) Number of Laye”
Mining Latent Structures for Multimedia Recommendation (2021, methodology): “There are three categories of GSL methods: metric learning [5, 21, 36], probabilistic modeling [7, 24, 43], and direct optimization approaches [8, 16, 41].”
A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects (2020, methodology): “[37] constructs a residual graph using a learnable distance function with two-node features as input.”
Simple and Deep Graph Convolutional Networks (2020, background): “The idea is that a feedforward neural network can be considered as an iterative optimization algorithm to minimize some function, and it was hypothesized that better optimization algorithms might lead to better network structure (Li et al., 2018a).”
DeeperGCN: All You Need to Train Deeper GCNs (2020, methodology): “omain. Later, many methods are proposed to apply improvements, extensions, and approximations on spectral-based GCNs [Kipf and Welling,2016,Defferrard et al.,2016,Henaff et al.,2015,Levie et al.,2018,Li et al., 2018]. On the other hand, spatial-based GCNs [Hamilton et al.,2017,Monti et al.,2017a,Nie”
GMAN: A Graph Multi-Attention Network for Traffic Prediction (2019, background): “A line of studies generalize CNN to model arbitrary graphs on spectral (Defferrard, Bresson, and Vandergheynst 2016; Kipf and Welling 2017; Li et al. 2018a) or spatial (Atwood and Towsley 2016; Hamilton, Ying, and Leskovec 2017; Chen, Ma, and Xiao 2018) perspective.”
NeVAE: A Deep Generative Model for Molecular Graphs (2018, background): “st of the approaches in the literature are based on matrix factorization [8, 4, 11], random walks [15, 34], or neighborhood aggregation functions [16], which include graph convolution networks (GCNs) [24, 39, 9, 30, 43]. Hamilton et al. [17] provides a comprehensive survey. However, this line of wor”
Designing Random Graph Models Using Variational Autoencoders With Applications to Chemical Design (2018, neutral): “Most of the approaches in the literature are based on matrix factorization [8, 4, 11], random walks [15, 34], or neighborhood aggregation functions [16], which include graph convolution networks (GCNs) [24, 39, 9, 30, 43].”
Diffusion Improves Graph Learning (2019, methodology): “In general, GNNs are classified into spectral-based models [11, 16, 28, 32, 41], which are based on the eigendecomposition of the graph Laplacian, and spatialbased methods [23, 26, 43, 52, 56, 62, 74], which use the graph directly and form new representations by aggregating the representations of a ”