Meta researchers release a much faster way to group data
Meta's FAIR lab has released e-means, a replacement for the widely used k-means method of sorting data into groups. The researchers say it matches 25 rounds of k-means in under three rounds. They report it runs 10 to 45 times faster on graphics chips. It is free in Meta's FAISS software library. The results have not been independently checked.
Where this came from. One outlet reported this. We read it, then wrote the sixty words above.