Deep Learning-Based Attribute Graph Clustering: An Overview

The application of deep learning in processing graph data has become increasingly important. In recent years, due to the powerful representation learning capabilities of deep learning, attribute graph clustering has emerged as a crucial method for dealing with complex network structures. In the field of network information security, a profound understanding and accurate classification of complex networks are particularly critical. In this article, we review some representative methods for deep graph clustering, including both classical and state-of-the-art approaches. These methods can be categorized into three types based on their principles: generative, adversarial, and contrastive. Generative methods employ deep learning models to generate data samples with attribute information, thereby improving clustering performance. Adversarial methods involve multiple competitive networks that interact to enhance clustering quality and model robustness. Contrastive methods improve clustering effectiveness by comparing the similarity or dissimilarity between different data points using similarity metrics. Finally, we point out several potential challenges and directions in the field of deep learning for attribute graph clustering. This work is expected to provide insights for researchers interested in this field and further advance the development of attribute graph clustering.