cuSLINK: Single-linkage Agglomerative Clustering on the GPU

TL;DR

cuSLINK introduces a GPU-optimized single-linkage clustering algorithm that significantly reduces space complexity and accelerates clustering tasks across various data mining applications.

Contribution

The paper presents cuSLINK, a novel GPU reformulation of SLINK with reduced space requirements and reusable computational primitives for efficient clustering.

Findings

Achieves state-of-the-art GPU clustering performance

Reduces space complexity to O(Nk)

Enables scalable clustering for large datasets

Abstract

In this paper, we propose cuSLINK, a novel and state-of-the-art reformulation of the SLINK algorithm on the GPU which requires only $O(Nk)$ space and uses a parameter $k$ to trade off space and time. We also propose a set of novel and reusable building blocks that compose cuSLINK. These building blocks include highly optimized computational patterns for $k$-NN graph construction, spanning trees, and dendrogram cluster extraction. We show how we used our primitives to implement cuSLINK end-to-end on the GPU, further enabling a wide range of real-world data mining and machine learning applications that were once intractable. In addition to being a primary computational bottleneck in the popular HDBSCAN algorithm, the impact of our end-to-end cuSLINK algorithm spans a large range of important applications, including cluster analysis in social and computer networks, natural language processing, and computer vision. Users can obtain cuSLINK at https://docs.rapids.ai/api/cuml/latest/api/#agglomerative-clustering