3 Commits

Author	SHA1	Message	Date
Matthijs Douze	0d7817e88f	rewrite python kmeans without scipy (#3873) ... Summary: Pull Request resolved: https://github.com/facebookresearch/faiss/pull/3873 The previous version required scipy to do the accumulation, which is replaced here with a nifty piece of numpy accumulation. This removes the need for scipy for non-sparse data. Reviewed By: junjieqi Differential Revision: D62884307 fbshipit-source-id: 5443634e487387a2b518fd2a7f9a3d9a40abd4b4	2024-09-20 09:15:27 -07:00
Matthijs Douze	016aa04602	make balanced clusters the default (#2796) ... Summary: Pull Request resolved: https://github.com/facebookresearch/faiss/pull/2796 This diff makes balanced clusters the default for 2-level clustering. This seems to improve a bit over the default uniform clusters, see https://github.com/fairinternal/faiss_improvements/blob/master/better_coarse_quantizer/two_level_clustering.ipynb Warning: the nc2 argument of two_level_clustering becomes the *total* number of clusters. Reviewed By: algoriddle Differential Revision: D44421222 fbshipit-source-id: 951b7fc043be4a41762a7e6f7a6fcfb71e303832	2023-03-28 07:23:30 -07:00
Matthijs Douze	b8fe92dfee	contrib clustering module (#2217) ... Summary: Pull Request resolved: https://github.com/facebookresearch/faiss/pull/2217 This diff introduces a new Faiss contrib module that contains: - generic k-means implemented in python (was in distributed_ondisk) - the two-level clustering code, including a simple function that runs it on a Faiss IVF index. - sparse clustering code (new) The main idea is that that code is often re-used so better have it in contrib. Reviewed By: beauby Differential Revision: D34170932 fbshipit-source-id: cc297cc56d241b5ef421500ed410d8e2be0f1b77	2022-02-28 14:18:47 -08:00