e-means beats k-means with 10x fewer epochs, ships in FAISS with order-of-magnitude speedups
npinto · x · 2026-11-11
Classic ML gets an upgrade: e-means is a new clustering algorithm claiming to beat k-means using 10x fewer epochs, with order-of-magnitude wall-clock speedups.
- Released in FAISS, no hyperparameter tuning needed beyond k
- Robust across various scales and datasets
- Author invites skeptics to try it themselves in an accompanying thread
More from Research
- Rubin Observatory opens data floodgates — a discovery wave is coming — mjuric · 2026-11-11
- Infrathon: 10,000 GPU-hours benchmark shows agents beat human infra baselines — UW · 2026-11-11
- Claude formalizes the full Ternary Goldbach proof in Lean 4, no sorry, three axioms — Singularitarian · 2026-11-11
- Gautam Kamath wins FOCS Test of Time Award for 2016 robust high-dimension estimators paper — thegautamkamath · 2026-11-11
- Ramez Naam: seminal takeoff models don't match empirical data, multi-agent scaling gains far flatter than predicted — ramez · 2026-11-11
- Metaculus AGI forecast jumps from 2033 to mid-2030 as forecasters turn radical — Afinetheorem · 2026-11-11