Adjust IVF fixup phase to sometimes bypass some of the neighborhood calculations #130490
+55
−49
Conversation
This file contains hidden or 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
elasticsearchmachine
added
the
Team:Search Relevance
|
Pinging @elastic/es-search-relevance (Team:Search Relevance) |
benwtrent
changed the title
iverase
reviewed
Jul 3, 2025
| neighborhoods.set(i, neighbors); | ||
| float maxIntraDistance = queue.consumeNodesWithWorstScore(neighbors, scores); | ||
| // Sort neighbors by their score | ||
| for (int j = 0; j < neighborCount; j++) { |
There was a problem hiding this comment.
Why not populate the array using the pop method of the priority queue?
for (int j = neighborCount - 1; j >= 0; j--) {
neighbors[j] = queue.pop();
}
There was a problem hiding this comment.
@iverase let me benchmark this
iverase
approved these changes
Jul 3, 2025
benwtrent
added
the
auto-merge-without-approval
iverase
reviewed
Jul 3, 2025
| continue; | ||
| } | ||
| // consume the queue into the neighbors array and get the maximum intra-cluster distance | ||
| int[] neighbors = new int[queue.size()]; |
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Outdated suggestions cannot be applied.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.
During the fixup phase, we compare the vector against every neighbor in the cluster neighborhood, no matter what. This seems pretty wasteful, especially for tightly clustered set of neighbors.
This change adjusts the fix up phase to only check for assignment if the currently assigned centroid is worse than the maximum intra-distance of the neighborhood.
This further reduces index time at no perceivable recall loss. I ran over 3 data sets, multi-segment and force merged.
Additionally, I noticed that we seemed to compute neighborhoods and use those calculations even when the total number of clusters is fewer than the neighborhood size. I adjusted this logic and we only compute the neighborhoods when the number of clusters is larger than the configured fixup neighborhood size.
All in all, this gives us about 5-15% index performance boost with no substantial drop in recall (the most I saw across all my runs was 0.01)
@iverase @john-wagster let me know what y'all think