Towards Robust and Scalable Density-based Clustering via Graph Propagation

2026-05-01

AI Summary

This work addresses the limitations of traditional density-based clustering methods—namely, their sensitivity to parameters and poor scalability in high-dimensional, large-scale data—by introducing a novel neighborhood-graph-based paradigm. The proposed approach formulates variable-density clustering as a label propagation process on a graph, effectively integrating graph connectivity with density mechanisms through a deterministic density propagation strategy. This integration bridges the theoretical gap between the two principles, substantially reducing parameter sensitivity while supporting arbitrary distance metrics. By combining efficient neighbor identification with fast propagation algorithms from network science, the method achieves minute-level clustering on datasets with millions of points, consistently outperforming existing baselines in accuracy.

Abstract

We present \textit{CluProp}, a novel framework that reimagines varied-density clustering in high-dimensional spaces as a label propagation process over neighborhood graphs. Our approach formally bridges the gap between density-based clustering and graph connectivity, leveraging efficient propagation mechanisms from network science to mitigate the parameter sensitivity inherent in traditional density-based methods. Specifically, we introduce a deterministic density-based propagation strategy to ensure scalable neighborhood identification. The framework is agnostic to the choice of distance metric and exhibits superior performance on large-scale data, processing millions of points in minutes while consistently outperforming existing baselines in accuracy.

Problem

• density-based clustering
• high-dimensional data
• parameter sensitivity
• scalability
• varied-density clustering

Innovation

• density-based clustering
• graph propagation
• label propagation
• scalable clustering
• parameter robustness