E-means: 10x faster k-means at equal loss

Researchers at Meta FAIR have introduced 'e-means,' a new clustering algorithm that significantly outperforms the traditional Lloyd's algorithm. It achieves a 10x speedup in both epochs and wall-clock time while maintaining lower loss metrics.
Why it matters
This advancement offers a more efficient way to handle large-scale data clustering, which is a fundamental operation in machine learning and data science workflows.
@misc{bessa2026emeansoutperformingkmeansin, title={e-means: outperforming k-means (in one-tenth the epochs)}, author={Swann Bessa and Pierre-Emmanuel Mazar and Gergely Szilvasy and Hugo Touvron and David Picard and Matthijs Douze and Herv J gou}, year={2026}, eprint={2611.09501}, archivePrefix={arXiv}, primaryClass={cs.LG}, url={https://arxiv.org/abs/2611.09501}, } Copy citation Open on arXiv Code e-means is available in faiss : faiss/contrib/e_means e-means: outperforming k-means in one-tenth the epochs I'm excited to open-source e-means: a modern, robust k-means algorithm developed at Meta FAIR. A replacement of Lloyd's algorithm, e-means delivers a ~10 speedup in both epochs and wall-clock time while achieving lower loss.
Pierre-Emmanuel Mazar , Gergely Szilvasy, Hugo Touvron, David Picard, Matthijs Douze and Herv J gou
100,000 datapoints in 200 clusters, two wide Gaussians and three small ones, cut off at a box, on a light uniform background. Lloyd has 1 update against 97 per epoch for e-means.
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in