e-means: A Faster k-means Variant and the Jevons Paradox of Cheaper Clustering
Start here
A new k-means variant called e-means updates centroids continuously within each pass using mini-batches and exponential moving averages, instead of the usual full-batch update. On 10 billion DINO embeddings clustered into roughly 16.7 million clusters, it reached the test loss of full-batch k-means in 2.5 epochs rather than 25, according to the author. The method keeps the same objective and convex nearest-centroid geometry, and fitted scaling laws set batch size and update rate automatically. The author argues that cheaper clustering runs will lead people to try more values of k rather than to lower compute bills.
Generated from reports · updated 13 h ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AIe-means: A Faster k-means Variant and the Jevons Paradox of Cheaper Clustering
e-means is a new k-means variant that updates centroids continuously within each pass using mini-batches and exponential moving averages, reaching the test loss of full-batch k-means in 2.5 epochs instead of 25 on 10 billion DINO embeddings clustered into roughly 16.7 million clusters. It keeps the same objective and convex nearest-centroid geometry, with fitted scaling laws setting batch size and update rate automatically. The author argues cheaper runs mean trying more values of k rather than lower compute bills.
Heat for this event
Not enough continuous observations yet to draw a trend.