3 ms·
> I’d still take HNSW over such approach Why? Sorry, but your comment is not very insightful without more explanation, and the rest of your comment is just an
by moab 3y ago
> I’d still take HNSW over such approach
Why? Sorry, but your comment is not very insightful without more explanation, and the rest of your comment is just an advertisement for your (admittedly impressive looking) work.
- ashvardanian 3y agoBecause (A) it allows similarity search between auxiliary objects, not just equidimensional vectors, and (B) can be implemented in a similarly short fashion, while providing great performance.