Abstract:Graph clustering, an essential unsupervised learning task in data mining, has garnered significant attention in recent years. With the advent of deep learning, considerable progress has been made in this field. However, existing methods present several limitations: (1) Most encoder models employ Graph Convolutional Networks (GCNs) as encoders. However, GCNs assign equal weight to each neighboring node and have been shown to be oversmoothing, thereby impacting clustering performance. (2) Most algorithms do not fully utilize the original graph content and structural information, leading to incomplete embedding features. (3) These methods do not account for the specific distribution of clustering of embedding features and the enhancement of staged pseudo-labels on clustering tasks.In this study,we propose a novel end-to-end graph clustering model that leverages graph attention encoders. Specifically, we initially employ a graph attention encoder to extract the inherent information from the original graph. This process assigns varying weights to different nodes, thereby avoiding excessive smoothing. We also fully utilize the guidance of periodic pseudo-labels to facilitate the learning of potential features that are beneficial for clustering. In addition, to improve the model's clustering performance, we introduce a regularization term that distributes the node features of different classifications across distinct low-dimensional spaces. Furthermore, to prevent the embedding features from straying from the original graph features, we design an information consistency module. Experimental results on the node graph datasets show that our model outperforms other state-of-the-art algorithms.

Unsupervised Feature Extraction Using a Learned Graph with Clustering Structure

Self-Adaptive Clustering of Dynamic Multi-Graph Learning

Clustering-Guided Sparse Structural Learning for Unsupervised Feature Selection

Learning Robust Graph for Clustering

Unsupervised Discriminative Feature Selection via Contrastive Graph Learning

Unsupervised discriminative feature learning via finding a clustering-friendly embedding space

The comparative recidivism rates of voluntary- and coerced-admission male alcoholics.

Structured graph learning for clustering and semi-supervised classification

Semi-Supervised Clustering via Dynamic Graph Structure Learning

Structured Optimal Graph-Based Clustering with Flexible Embedding.

Cluster-aware Multiplex InfoMax for Unsupervised Graph Representation Learning

Self-Supervised Graph Embedding Clustering

Unsupervised feature selection via multiple graph fusion and feature weight learning

Efficient and Stable Unsupervised Feature Selection Based on Novel Structured Graph and Data Discrepancy Learning

Self-supervised graph clustering via attention auto-encoder with distribution specificity

Unsupervised feature selection with high-order similarity learning

Towards attributed graph clustering using enhanced graph and reconstructed graph structure

Unsupervised Spectral Feature Selection with Dynamic Hyper-graph Learning

Adaptive and structured graph learning for semi-supervised clustering

Unsupervised Dimensionality Reduction Based on Fusing Multiple Clustering Results

Unsupervised Feature Selection Algorithm Based on Graph Filtering and Self-representation