Attributed Graph Clustering: A Deep Attentional Embedding Approach

Wang, Chun; Pan, Shirui; Hu, Ruiqi; Long, Guodong; Jiang, Jing; Zhang, Chengqi

doi:10.48550/arxiv.1906.06532

Cited by 30 publications

(53 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is because both SDCN and DFCN overly introduce the attribute information learned by the auto-encoder part into the latent space, so that the node embedding contains redundant attributes about the sample, leading to representation collapse. In contrast, by reducing the information correlation in a dual manner, DCRN can learn more meaningful representation to improve the clustering performance; 2) it can be observed that the GCNbased clustering methods GAE/VGAE (Kipf and Welling 2016b), ARGA (Pan et al 2019) and DAEGC (Wang et al 2019) are not comparable with ours. This is because these methods do not consider to handle information correlation redundancy, thus resulting in the trivial constant representation; 3) our method improves the auto-encoder-based clustering methods, i.e., AE (Yang et al 2017), DEC (Yang et al 2017) and IDEC (Guo et al 2017), by a large margin, all of which have been verified strong representation learning capacity for clustering on non-graph data, while these methods that merely rely on attribute information can not effec-tively learn discriminative information on graphs; 4) since K-means (Hartigan and Wong 1979) is directly performed on raw attributes, thus achieving unpromising results.…”

Section: Performance Comparisonmentioning

confidence: 67%

“…Specifically, GAE/VGAE (Kipf and Welling 2016b) embeds the node attributes with structure information via a graph encoder and then reconstructs the graph structure by an inner product decoder. Inspired by their success, recent researches, DAEGC (Wang et al 2019), GALA (Park et al 2019), ARGA (Pan et al 2019) and AGAE (Tao et al 2019) further improve the early works with graph attention network, Laplacian sharpening, and generative adversarial learning. Although achieving promising clustering performance, the over-smoothing problem has not been effectively tackled in these methods, which affects the clustering performance.…”

Section: Related Work Attributed Graph Clusteringmentioning

confidence: 99%

“…Three representative deep generative methods, i.e., AE (Yang et al 2017), DEC (Xie, Girshick, and Farhadi 2016), and IDEC (Guo et al 2017), train an auto-encoder and then perform a clustering algorithm over the learned latent embedding. GAE/VGAE (Kipf and Welling 2016b), DAEGC (Wang et al 2019), and ARGA/ARVGA (Pan et al 2019) are three typical GCNbased frameworks that learn the representation for clustering by considering both node attribute and structure information. Furthermore, we report the performance of three stateof-the-art deep clustering methods, i.e., SDCN/SDCN Q (Bo et al 2020), DFCN (Tu et al 2020), and MVGRL (Hassani and Khasahmadi 2020), which utilize two sub-networks to process augmented graphs independently.…”

Section: Performance Comparisonmentioning

confidence: 99%

“…Thanks to the powerful graph information exploitation capability, graph convolutional networks (GCN) (Kipf and Welling 2016a) have recently achieved promising performance in many graph clustering applications like social networks and recommendation systems. Consequently, it has attracted considerable attention in this field and many algorithms are proposed (Wang et al 2019;Pan et al 2019;Tao et al 2019;Park et al 2019;Bo et al 2020;Tu et al 2020). The heat maps of node similarity matrices in the latent space of GAE (Kipf and Welling 2016b), MVGRL (Hassani and Khasahmadi 2020), and our proposed method on the ACM dataset.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Deep Graph Clustering via Dual Correlation Reduction

Liu¹,

Tu²,

Zhou³

et al. 2021

Preprint

View full text Add to dashboard Cite

Deep graph clustering, which aims to reveal the underlying graph structure and divide the nodes into different groups, has attracted intensive attention in recent years. However, we observe that, in the process of node encoding, existing methods suffer from representation collapse which tends to map all data into the same representation. Consequently, the discriminative capability of the node representation is limited, leading to unsatisfied clustering performance. To address this issue, we propose a novel self-supervised deep graph clustering method termed Dual Correlation Reduction Network (DCRN) by reducing information correlation in a dual manner. Specifically, in our method, we first design a siamese network to encode samples. Then by forcing the cross-view sample correlation matrix and cross-view feature correlation matrix to approximate two identity matrices, respectively, we reduce the information correlation in the dual-level, thus improving the discriminative capability of the resulting features. Moreover, in order to alleviate representation collapse caused by over-smoothing in GCN, we introduce a propagation regularization term to enable the network to gain long-distance information with the shallow network structure. Extensive experimental results on six benchmark datasets demonstrate the effectiveness of the proposed DCRN against the existing state-of-the-art methods. The code of DCRN is available at DCRN and a collection of deep graph clustering is shared at Awesome Deep Graph Clustering on Github.

show abstract

Section: Performance Comparisonmentioning

confidence: 67%

Section: Related Work Attributed Graph Clusteringmentioning

confidence: 99%

Section: Performance Comparisonmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Deep Graph Clustering via Dual Correlation Reduction

Liu¹,

Tu²,

Zhou³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Graph clustering has also been used to improve the training for graph encoders. DAEGC(Wang et al 2019a) and SDCN(Bo et al 2020) jointly optimize clustering algorithms and the graph reconstruction loss. AGC(Zhang et al 2019) adaptively finds the optimal order for GCN based on the intrinsic clustering scores.…”

mentioning

confidence: 99%

X-GOAL: Multiplex Heterogeneous Graph Prototypical Contrastive Learning

Jing,

Feng,

Xiang

et al. 2021

Preprint

View full text Add to dashboard Cite

Contrastive Learning (CL) is one of the most popular selfsupervised learning frameworks for graph representation learning, which trains a Graph Neural Network (GNN) by discriminating positive and negative node pairs. However, there are two challenges for CL on graphs. On the one hand, traditional CL methods will unavoidably introduce semantic errors since they will treat some semantically similar nodes as negative pairs. On the other hand, most of the existing CL methods ignore the multiplexity nature of the real-world graphs, where nodes are connected by various relations and each relation represents a view of the graph. To address these challenges, we propose a novel Graph Multi-View Prototypical (Graph-MVP) framework to extract node embeddings on multiplex graphs. Firstly, we introduce a Graph Prototypical Contrastive Learning (Graph-PCL) framework to capture both node-level and semantic-level information for each view of multiplex graphs. Graph-PCL captures the nodelevel information by a simple yet effective data transformation technique. It captures the semantic-level information by an Expectation-Maximization (EM) algorithm, which alternatively performs clustering over node embeddings and parameter updating for GNN. Next, we introduce Graph-MVP based on Graph-PCL to jointly model different views of the multiplex graphs. Our key insight behind Graph-MVP is that different view-specific embeddings of the same node should have similar underlying semantic, based on which we propose two versions of Graph-MVP: Graph-MVP hard and Graph-MVP sof t to align embeddings across views. Finally, we evaluate the proposed Graph-PCL and Graph-MVP on a variety of real-world datasets and downstream tasks. The experimental results demonstrate the effectiveness of the proposed Graph-PCL and Graph-MVP frameworks.

show abstract

Data Clustering and Representation Learning Based on Networked Data

Labiod,

Nadif

2023

Studies in Classification, Data Analysis, and Knowledge Organization

View full text Add to dashboard Cite

To deal simultaneously with both, the attributed network embedding and clustering, we propose a new model exploiting both content and structure information. The proposed model relies on the approximation of the relaxed continuous embedding solution by the true discrete clustering. Thereby, we show that incorporating an embedding representation provides simpler and easier interpretable solutions. Experiment results demonstrate that the proposed algorithm performs better, in terms of clustering, than the state-of-art algorithms, including deep learning methods devoted to similar tasks.

show abstract

Attributed Graph Clustering: A Deep Attentional Embedding Approach

Cited by 30 publications

References 12 publications

Deep Graph Clustering via Dual Correlation Reduction

Deep Graph Clustering via Dual Correlation Reduction

X-GOAL: Multiplex Heterogeneous Graph Prototypical Contrastive Learning

Data Clustering and Representation Learning Based on Networked Data

Contact Info

Product

Resources

About