Binary Vector Search (with KNN reranking for recall accuracy boost) #709
MentalGear
started this conversation in
Ideas
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I've been following Orama for quite some time now. That's why I want to first say a great thank you to the whole team to provide this amazing universal and local search solution! The hybrid search is also a very unique feature.
Now, I was wondering if you already considered providing the option to enable a binary index for vector search?
This could boost performance 10x or more, which is especially interesting on local/edge devices or servers with millions of records.
In this regard, this article describes an interesting technique that basically uses a binary index for a wide retrieval followed by a KNN re-ranking to boost accuracy ( up to 95% ).
Binary Vector Search: The 30x Memory Reduction Revolution with Preserved Accuracy
Also, the HN comments on the article provide some more interesting methods.
Let me know what your thoughts on this are.
Beta Was this translation helpful? Give feedback.
All reactions