e-means surpasses k-means
The story so far
Meta FAIR released the e-means algorithm, achieving k-means 25 iterations' loss level within 1.4 to 2.6 iterations without hyperparameter tuning, with a 10-45x speedup on GPU and an 8-15x speedup on CPU, and has been integrated into FAISS 1.16.0. e-means achieves dual optimization of speed and loss through exponential moving average updates, automatic hyperparameter adjustment, improved initialization, and dead cluster revival strategies. Earlier reports stated that e-means achieved test loss lower than Lloyd's algorithm within 25 rounds of training, requiring only 1.2 to 3 rounds, with a 10-45x speedup on GPU and an 8-15x speedup on CPU compared to the state-of-the-art Flash-KMeans.
AI-written from the reports · Updated 8 days ago
Coverage timeline
Follow the coverage to see every side of the story.
- MarkTechPostMeta FAIR Releases e-means: An Algorithm Beating k-means in 10x Fewer Epochs
Meta FAIR's e-means algorithm achieves k-means's 25-iteration loss level within 1.4 to 2.6 iterations, with GPU speed improvements of 10-45 times and CPU speed improvements of 8-15 times, without hyperparameter tuning, and has been integrated into FAISS 1.16.0
- Hacker News · 300+E-means: Outperforming k-means in one-tenth the epochs
e-means achieved a test loss lower than Lloyd's algorithm within 25 rounds of training, requiring only 1.2 to 3 rounds. Compared to the state-of-the-art Flash-KMeans, GPU end-to-end speed improved 10-45 times, and CPU end-to-end speed improved 8-15 times. e-means achieved dual optimization of speed and loss through exponential moving average updates, automatic hyperparameter adjustment, improved initialization, and dead cluster revival strategies.
Heat over time
- Now
- 3
- Peak
- 512 Nov, 06:00
- 24h change
- -50%
Hover or tap the chart to see each hour; use the arrow keys on a keyboard.