Self-Supervised Graph Embedding Clustering

TL;DR

This paper introduces a self-supervised graph embedding framework that unifies manifold learning with K-means clustering, eliminating hyperparameters and maintaining class balance through a centroid-free approach, leading to improved clustering performance.

Contribution

It proposes a novel centroid-free K-means method integrated with manifold learning, ensuring one-step clustering without hyperparameters and theoretically maintaining class balance.

Findings

Effective clustering on multiple datasets

Elimination of hyperparameters in clustering process

Theoretical proof of class balance maintenance

Abstract

The K-means one-step dimensionality reduction clustering method has made some progress in addressing the curse of dimensionality in clustering tasks. However, it combines the K-means clustering and dimensionality reduction processes for optimization, leading to limitations in the clustering effect due to the introduced hyperparameters and the initialization of clustering centers. Moreover, maintaining class balance during clustering remains challenging. To overcome these issues, we propose a unified framework that integrates manifold learning with K-means, resulting in the self-supervised graph embedding framework. Specifically, we establish a connection between K-means and the manifold structure, allowing us to perform K-means without explicitly defining centroids. Additionally, we use this centroid-free K-means to generate labels in low-dimensional space and subsequently utilize the label information to determine the similarity between samples. This approach ensures consistency between the manifold structure and the labels. Our model effectively achieves one-step clustering without the need for redundant balancing hyperparameters. Notably, we have discovered that maximizing the $\ell_{2,1}$-norm naturally maintains class balance during clustering, a result that we have theoretically proven. Finally, experiments on multiple datasets demonstrate that the clustering results of Our-LPP and Our-MFA exhibit excellent and reliable performance.