Synergistic Deep Graph Clustering Network

TL;DR

A graph clustering framework named Synergistic Deep Graph Clustering Network (SynC), which designs a Transform Input Graph Auto-Encoder to obtain high-quality embeddings for guiding structure augmentation and introduces a structure fine-tuning strategy to improve the model's generalization.

Abstract

Employing graph neural networks (GNNs) to learn cohesive and discriminative node representations for clustering has shown promising results in deep graph clustering. However, existing methods disregard the reciprocal relationship between representation learning and structure augmentation. This study suggests that enhancing embedding and structure synergistically becomes imperative for GNNs to unleash their potential in deep graph clustering. A reliable structure promotes obtaining more cohesive node representations, while high-quality node representations can guide the augmentation of the structure, enhancing structural reliability in return. Moreover, the generalization ability of existing GNNs-based models is relatively poor. While they perform well on graphs with high homogeneity, they perform poorly on graphs with low homogeneity. To this end, we propose a graph clustering framework named Synergistic Deep Graph Clustering Network (SynC). In our approach, we design a Transform Input Graph Auto-Encoder (TIGAE) to obtain high-quality embeddings for guiding structure augmentation. Then, we re-capture neighborhood representations on the augmented graph to obtain clustering-friendly embeddings and conduct self-supervised clustering. Notably, representation learning and structure augmentation share weights, significantly reducing the number of model parameters. Additionally, we introduce a structure fine-tuning strategy to improve the model's generalization. Extensive experiments on benchmark datasets demonstrate the superiority and effectiveness of our method. The code is released on GitHub and Code Ocean.

Figures (8)

• Figure 1: Comparison of nodes similarity on the dataset ACM. The first row represents GAE, and the second represents GAE with linear transformation applied to $\mathbf{X}$.
• Figure 2: Illustration of our proposed SynC framework. The TIGAE combines linear transformation with graph convolutional networks. Firstly, we employ the pre-trained TIGAE to generate a predicted graph without a gradient. Secondly, in the fine-tuning phase, we apply the structure fine-tuning strategy to prune, connect, and assign weights to edges in the predicted graph. Subsequently, the refined graph is fed into TIGAE to learn information from neighbor nodes and obtain the final embeddings with a gradient. This is the synergistic interaction of representation learning and structure augmentation since the two TIGAE modules share weights. Finally, we conduct self-supervised clustering in the clustering phase using more cohesive and stronger discriminative embeddings.
• Figure 3: Ablation results of the proposed SynC on four datasets.
• Figure 4: Clustering results with different hyper-parameters.
• Figure 5: Convergence effect on different datasets.