Verified Evidence: resolved · Fact Lock: locked · QA: pass
Meta FAIR’s e-means algorithm has been shipped in FAISS 1.16.0 as a replacement for mini-batch k-means, according to MarkTechPost.

MarkTechPost reports that e-means reaches the loss achieved by Lloyd’s algorithm after 25 epochs in 1.4 to 2.6 epochs. The same report says the method delivers GPU speedups ranging from 10x to 45x and requires no additional hyperparameters.

Those performance figures remain single-source claims. MarkTechPost’s summary does not specify the datasets, hardware or experimental conditions behind the loss comparison and GPU speedup range, so the results cannot be independently assessed from the supplied evidence.

The available evidence establishes the reported inclusion of e-means in FAISS 1.16.0, but it does not independently establish the exact release date for either the algorithm or the software version.

SOURCES

  1. MarkTechPostMeta FAIR Releases e-means: An Algorithm Beating k-means in 10x Fewer Epochs