Efficient Sparse Spherical k-Means for Document Clustering

TL;DR

This paper introduces an optimized indexing method for spherical k-Means clustering that leverages data sparsity and convergence properties to enhance scalability for large cluster counts in document collections.

Contribution

It presents a novel indexing structure that significantly reduces comparison operations in spherical k-Means, improving its efficiency for large-scale document clustering.

Findings

Reduces the number of comparisons per iteration

Improves scalability with respect to the number of clusters

Maintains clustering quality while increasing efficiency

Abstract

Spherical k-Means is frequently used to cluster document collections because it performs reasonably well in many settings and is computationally efficient. However, the time complexity increases linearly with the number of clusters k, which limits the suitability of the algorithm for larger values of k depending on the size of the collection. Optimizations targeted at the Euclidean k-Means algorithm largely do not apply because the cosine distance is not a metric. We therefore propose an efficient indexing structure to improve the scalability of Spherical k-Means with respect to k. Our approach exploits the sparsity of the input vectors and the convergence behavior of k-Means to reduce the number of comparisons on each iteration significantly.