Oriel — AI & technology worth noticing, every day.

e-means: outperforming k-means (in one-tenth the epochs)

Paper · https://arxiv.org/abs/2611.09501

Lloyd's k-means algorithm is a long-running standard for clustering, with applications in data analysis, retrieval, quantization and data curation. In this paper, we introduce an exponential moving average optimization procedure for k-means with mini-batch updates, powered by modern deep learning machinery and new insights. We also establish scaling laws on how decay rate and batch size impact final performance. This gives us a formulaic rule to fix them for multiple problem sizes and embedding datasets, without the need for any hyperparameter sweep. The resulting clustering algorithm, coined e-means, is much faster than Lloyd k-means and converges to lower losses, frequently even beating k-means++. On a wide array of embeddings datasets, it beats the loss of Lloyd k-means at 25 epochs in 1.4 to 2.6 epochs, which translates to 10-45x GPU wall-clock speedups compared to state-of-the-art k-means implementations. As an example, e-means partitions DINO10B, a dataset of 10 billion DINO embeddings, into 16 million clusters in only 2.5 epochs, reaching a loss below Lloyd's at 25 epochs.

First seen by Oriel on .

AI summary

The paper introduces e-means, an exponential moving average mini-batch clustering algorithm that achieves lower losses than Lloyd's k-means while requiring substantially fewer epochs.

  • e-means uses modern deep learning machinery and scaling laws relating decay rate and batch size to final performance.
  • Its formulaic parameter rule works across multiple problem sizes and embedding datasets without hyperparameter sweeps.
  • Across embedding datasets, e-means beats Lloyd k-means at 25 epochs within 1.4 to 2.6 epochs, delivering 10-45x GPU wall-clock speedups.
  • e-means partitions 10 billion DINO embeddings into 16 million clusters in 2.5 epochs, reaching a lower loss than Lloyd's method at 25 epochs.

#clustering #k-means #embeddings #gpu computing

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Hugging Face Papers: 17, latest observation ; +12 over observed period since

Comments

Comment count reported by the source.

  • Hugging Face Papers: 3, latest observation ; 0 over observed period since

Aliases

SourceExternal IDURL
Hugging Face Papers2611.09501https://arxiv.org/abs/2611.09501 Read discussion