Similarity Search within the dataset

[Vikasdubey0551]

Hello,
I am new to the field of similarity search and have a quick question.
Is there a way/suggestion to obtain buckets containing similar vectors (Let's say with indexLSH) based on some distance criteria within the dataset instead of passing a query vector and doing an exhaustive search?

Thanks!

[mdouze]

Do you want to cluster the dataset?

[Vikasdubey0551]

@mdouze Thanks for the reply. Yeah. It definitely sounds like I need some clustering techniques like DBSCAN. However, they are slow and memory intensive so far in my attempts. The speed of faiss and GPU support is really attractive. I am not necessarily looking for extremely accurate results, Just more in line or perhaps a little better than Regular LSH. Any suggestions in that regard?

[qmpzzpmq]

Hi, @Vikasdubey0551 did you find anymethod to do with DBSCAN?