Discriminative Attribute Graph Clustering Through Topology-Guided Contrastive Learning

Deep attribute graph clustering aims to learn discriminative node representations by leveraging both node attributes and graph topology to partition nodes into distinct clusters. Although substantial progress has been made in attribute-graph clustering in recent years, two key challenges remain: noisy edges in the original adjacency matrix degrade the quality of information propagation, and redundant feature information across different feature views hampers the learning of discriminative representations. To address these issues, we propose a self-supervised attribute graph clustering method based on topological reconstruction and correlation decorrelation. First, we reconstruct the graph topology by computing intersections between k-nearest neighbors and the original adjacency relationships, while simultaneously leveraging global semantic information from K-means clustering to filter out noisy nodes. This reconstructed topology effectively mitigates information redundancy during feature aggregation in Graph Neural Networks. Second, unlike existing augmentation-based contrastive methods, we treat the feature representations from an auto-encoder (AE) and a graph auto-encoder (GAE) as two complementary natural views. We then apply mutual information minimization and a decorrelation constraint to suppress redundant information between views, yielding more discriminative node representations. Extensive experiments on four widely-used graph datasets—ACM, DBLP, CITE, and AMAP—demonstrate that our method consistently outperforms six state-of-the-art baselines.

Many real-world data can be represented as graphs, such as citation networks, social networks, and recommendation systems. In these graphs, each node usually has its own attributes, and nodes are connected by edges. Attribute graph clustering aims to automatically group similar nodes without using manual labels.However, graph data often contains noisy or unreliable connections, which can mislead the model when learning node representations. In addition, different types of learned features may contain overlapping or redundant information, making it harder to form clear clusters. To address these problems, this paper proposes a self-supervised graph clustering method. The method first reconstructs the graph structure by combining local neighbor information with global semantic information, helping the model reduce the influence of noisy edges. It then uses two neural network models, an auto-encoder and a graph auto-encoder, to learn complementary views of each node. By reducing redundant information between these views, the method learns more useful and distinguishable node representations. Experiments on four common graph datasets show that the proposed method improves clustering performance compared with several existing methods. This suggests that cleaner graph structures and better use of complementary feature views can help machines group graph data more accurately.