Improving Spectral Clustering Using Spectrum-Preserving Node Aggregation

TL;DR

This paper introduces a spectrum-preserving node aggregation method to accelerate spectral clustering on large datasets, significantly improving efficiency and accuracy by reducing eigen-decomposition complexity.

Contribution

The paper proposes a novel spectrum-preserving node reduction technique that enables nearly-linear time spectral clustering with enhanced accuracy on large-scale data.

Findings

Clustering performance is significantly improved over state-of-the-art methods.

The method reduces computational time to nearly-linear scale.

Experimental results demonstrate effective data representation and clustering accuracy.

Abstract

Spectral clustering is one of the most popular clustering methods. However, the high computational cost due to the involved eigen-decomposition procedure can immediately hinder its applications in large-scale tasks. In this paper we use spectrum-preserving node reduction to accelerate eigen-decomposition and generate concise representations of data sets. Specifically, we create a small number of pseudonodes based on spectral similarity. Then, standard spectral clustering algorithm is performed on the smaller node set. Finally, each data point in the original data set is assigned to the cluster as its representative pseudo-node. The proposed framework run in nearly-linear time. Meanwhile, the clustering accuracy can be significantly improved by mining concise representations. The experimental results show dramatically improved clustering performance when compared with state-of-the-art methods.