Subspace Clustering Without Knowing the Number of Clusters: A Parameter Free Approach

By Vishnu Menon et al

Published on June 20, 2020

Read the original document by opening this link in a new tab.

Table of Contents

I. Introduction
II. Overview and Essential Definitions
III. Algorithm
IV. Analysis
V. Numerical Results
VI. Discussion

Summary

Subspace clustering is a fundamental task in unsupervised machine learning that aims to cluster high dimensional data points coming from a union of subspaces. Existing algorithms often require prior knowledge of the number of clusters, but this paper proposes a parameter free approach. The algorithm focuses on the statistical distributions of angles subtended by data points within and across subspaces. By merging clusters based on the Bhattacharyya distance between angle distributions, the algorithm achieves clustering without predefining the number of clusters. The performance of the proposed method is compared with state-of-the-art techniques on synthetic and real datasets, demonstrating its effectiveness and lower computational complexity.