Stop exploring HNSW graph if scores are not getting better. #12770
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
benwtrent
commented
Nov 6, 2023
benwtrent
changed the title
|
oh, good catch! I guess equality is unusual, but this should help in some cases. I wonder if it helps with degenerate case where all scores are equal? EG all zero vectors (see #11626) |
It would help by ending exploration for sure. In this degenerate case, it is even worse than causing slowness. If there are millions of vectors we could cause an OOME as there are no barriers preventing the candidate list from expanding (we only pop one, but then could add 15 more on each neighbor explored for example). |
mayya-sharipova
approved these changes
Nov 6, 2023
msokolov
approved these changes
Nov 6, 2023
benwtrent
added a commit
that referenced
this pull request
Nov 7, 2023
I noticed while testing lower dimensionality and quantization, we would explore the HNSW graph way too much. I was stuck figuring out why until I noticed the searcher checks for distance equality (not just if the distance is better) when exploring neighbors-of-neighbors. This seems like a bad heuristic, but to double check I looked at what nmslib does. This pointed me back to this commit: nmslib/nmslib#106 Seems like this performance hitch was discovered awhile ago :). This commit adjusts HNSW to only explore the graph layer if the distance is actually better.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I noticed while testing lower dimensionality and quantization, we would explore the HNSW graph way too much. I was stuck figuring out why until I noticed the searcher checks for distance equality (not just if the distance is better) when exploring neighbors-of-neighbors. This seems like a bad heurstic, but to double check I looked at what nmslib does. This pointed me back to this commit: nmslib/nmslib#106
Seems like this performance hitch was discovered awhile ago :).
This commit adjusts HNSW to only explore the graph layer if the distance is actually better.