Dual Graph Convolutional Networks for Graph-Based Semi-Supervised Classification

By Chenyi Zhuang, Qiang Ma. The Web Conference, 2018

441 citations recorded by the source corpus.

The problem of extracting meaningful data through graph analysis spans a range of different fields, such as the internet, social networks, biological networks, and many others. The importance of being able to effectively mine and learn from such data continues to grow as more and more structured data become available. In this paper, we present a simple and scalable semi-supervised learning method for graph-structured data in which only a very small portion of the training data are labeled. To sufficiently embed the graph knowledge, our method performs graph convolution from different views of the raw data. In particular, a dual graph convolutional neural network method is devised to jointly consider the two essential assumptions of semi-supervised learning: (1) local consistency and (2) global consistency. Accordingly, two convolutional neural networks are devised to embed the local-consistency-based and global-consistency-based knowledge, respectively. Given the different data transformations from the two networks, we then introduce an unsupervised temporal loss function for the ensemble. In experiments using both unsupervised and supervised loss functions, our method outperforms state-of-the-art techniques on different datasets.

Subfields: Artificial Intelligence, Statistical and Nonlinear Physics

View Paper: https://www.semanticscholar.org/paper/d0b66e82f9386c16c489fb5e4bb5fa01a2da754b

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 Statistical and Nonlinear Physics

Citations in the corpus

Knowledge Graph Contrastive Learning for Recommendation (2022, methodology): “…knowledge graph structures, we obtain two knowledge graph dependency views with operators Inspired by the investigation of graph consistency in [16, 55], to explore the agreement property of each item based on the augmented views, we define knowledge graph structure consistency 𝑐 𝑖 of item 𝑖…”
Graph Learning: A Survey (2021, neutral): “DGCN uses two convolutions: one is the characteristic form of polynomial filter, and the other is to replace the adjacency matrix with the PPMI (Positive Pointwise Mutual Information) of the transition probability [89].”
Hypergraph Convolution and Hypergraph Attention (2019, background): “(Zhuang & Ma, 2018) further explore a joint usage of diffusion and adjacency basis in a dual graph convolutional network.”
Dual Graph Attention Networks for Deep Latent Representation of Multifaceted Social Effects in Recommender Systems (2019, background): “[45] extends GCN to dual structures, combining global and local consistencies in graph.”
A Survey on Deep Semi-Supervised Learning (2021, background): “Subsequently, GCN extensions for SSL started to propagate [193] [194] [195] [196] [197].”
The Klarna Product Page Dataset: A Realistic Benchmark for Web Representation Learning (2021, background): “GCNs have shown notable performance on social network graphs [39, 6, 5], citation networks [43, 9], and biochemical graphs [6, 39, 24].”
A Comprehensive Survey on Graph Neural Networks (2019, background): “…until a stable equilibrium is Recurrent Graph Neural Networks (RecGNNs) [15], [16], [17], [18] Spectral methods [19], [20], [21], [22], [23], [40], [41] Convolutional Graph Neural Networks (ConvGNNs) Spatial methods [24], [25], [26], [27], [42], [43], [44] [45], [46], [47], [48], [49], [50], [51]…”
A Review of Graph Neural Networks and Their Applications in Power Systems (2021, background): “Furthermore, a dual-graph convolutional network (DGCN) that consists of two graph convolutional layers in parallel is proposed to jointly account for the global consistency and lo‐ cal consistency in [32].”
Reliable Data Distillation on Graph Convolutional Network (2020, methodology): “We compare the performances of the last single model in RDD with other start-of-the-art non-ensemble models: GAT [48], GPNN [36], APPNP [19], LGCN [22], NGCN [1], and DGCN [64].”
AWB-GCN: A Graph Convolutional Network Accelerator with Runtime Workload Rebalancing (2019, methodology): “The GCN model conﬁguration follows the original GCN algorithm papers [29], [46], [47].”
Label Efficient Semi-Supervised Learning via Graph Filtering (2019, background): “Other related works include GraphSAGE [25], graph attention networks [51], attention-based graph neural network [47], graph partition neural networks [36], FastGCN [16], dual graph convolutional neural network [65], stochastic GCN [17], Bayesian GCN [58], deep graph infomax [50], LanczosNet [37], et”
A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects (2020, methodology): “Zhuang and Ma [38] implemented a dual-GCN (DGCN) by parallelizing two graph convolution layers.”
Deep Learning on Graphs: A Survey (2018, positive): “DGCN [51] was further proposed to jointly adopt the diffusion and the adjacency bases using a dual graph convolutional network.”
Graph convolutional networks: a comprehensive review (2019, background): “Many graph convolutional network models have been proposed, to name a few, including [5, 37, 42, 61, 95].”
Adaptively Connected Neural Networks (2019, methodology): “MoNet [30], GAT [41], LGCN [8], and Dual GCN [59] are used as competing methods.”
Graph Neural Networks: A Review of Methods and Applications (2018, background): “The dual graph convolutional network (DGCN) (Zhuang and Ma, 2018) is proposed to jointly consider the local consistency and global consistency on graphs.”
Unsupervised Attributed Multiplex Network Embedding (2019, background): “…0.798 0.412 0.825 network structure; CAN (Meng et al. 2019): It learns embeddings of both attributes and nodes in the same semantic space; DGCN (Zhuang and Ma 2018): It models the local and global properties of a graph by employing dual GCNs. 2) Multiplex embedding methods • No attributes:…”