An effective variant of the Hartigan $k$-means algorithm

2026-04-23 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, quick

Summary

A minor variation of Hartigan's k-means algorithm has been developed, yielding an additional 2% to 5% improvement over the original method. Hartigan's algorithm (1975) already outperforms Lloyd's algorithm (1957) by 5% to 10% in most clustering scenarios, making this new variant a significant advancement. The performance gains of this improved Hartigan variant tend to increase with higher data dimensionality or a larger number of clusters (k). This development offers a more effective approach to the classical k-means clustering problem.

Key takeaway

For data scientists and machine learning engineers working with clustering problems, adopting this improved Hartigan k-means variant can lead to more accurate and robust results. If your applications involve high-dimensional data or a large number of clusters, you should prioritize evaluating this variant to achieve a 2% to 5% performance gain over standard Hartigan implementations.

Key insights

A minor Hartigan k-means variant improves clustering results by an additional 2-5%.

Principles

• Hartigan's algorithm outperforms Lloyd's.
• Improvements scale with dimension or k.

Method

The method involves a very minor variation to Hartigan's k-means algorithm, enhancing its clustering performance, particularly in higher dimensions or with more clusters.

In practice

• Apply variant for better clustering.
• Consider for high-dimensional data.

Topics

• k-means Algorithm
• Hartigan's Algorithm
• Lloyd's Algorithm
• Clustering Algorithms
• Algorithm Variants

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Data Scientist

Related on AIssential

• k-means #maths #dataanlysis #datascience #machinelearning — K-means clustering iteratively refines data point assignments to centroids to discover inherent data groupings.
• Adaptive Weighted Averaging — The Adaptive Weighted Averaging strategy identifies the largest value among n options, guaranteeing performance never worse than baselines.
• Comparison of Outlier Detection Algorithms on String Data — Two novel algorithms effectively detect outliers in string data, a less-explored area in machine learning.
• A Data-Informed Variational Clustering Framework for Noisy High-Dimensional Data — DIVI is a variational clustering framework that jointly learns feature relevance and adapts cluster structure in noisy high-dimensional data.
• CLARITree: Cholesky and Lookahead Accelerations for Regression with Interpretable Piecewise Linear Trees — CLARITree combines lookahead search and Cholesky updates for efficient, near-optimal, sparse piecewise linear regression trees.
• DAG-MoE: From Simple Mixture to Structural Aggregation in Mixture-of-Experts — DAG-MoE improves Mixture-of-Experts performance by replacing weighted-summation with learned structural aggregation, expanding expert-combination space and enabling multi-step reasoning.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.