English

Simple KNN-Based Outlier Detection Achieves Robust Clustering

Machine Learning 2026-05-11 v1 Data Structures and Algorithms

Abstract

Being robust to the presence of outliers is crucial for applying clustering algorithms in practice. In the \textit{robust k-Means} problem (i.e., kk-Means with outliers), the goal is to remove zz outliers and minimize the kk-Means cost on the remaining points. Despite the close connection between robust kk-Means and outlier detection, both theoretical and empirical understanding of the effectiveness of classic outlier detection heuristics\textit{classic outlier detection heuristics} for robust kk-Means remains limited. In this paper, we prove that under a practical assumption on the optimal cluster sizes, simply removing points with large KK-Nearest-Neighbor distances achieves performance comparable to prior work in terms of approximation guarantees: it yields a constant-factor reduction from robust kk-Means to standard kk-Means, without introducing additional centers or discarding extra outliers, as is commonly required by existing approaches. Empirically, experiments on real-world datasets show that our method outperforms or matches several more sophisticated algorithms in terms of clustering cost and runtime. These results demonstrate that simple KNN-based heuristics can be surprisingly effective for robust clustering, highlighting new opportunities to bridge techniques from outlier detection and clustering.

Keywords

Cite

@article{arxiv.2605.07130,
  title  = {Simple KNN-Based Outlier Detection Achieves Robust Clustering},
  author = {Tianle Jiang and Yufa Zhou},
  journal= {arXiv preprint arXiv:2605.07130},
  year   = {2026}
}

Comments

Code: https://github.com/MasterZhou1/Robust-Clustering

R2 v1 2026-07-01T12:56:42.834Z