jpountz opened a new issue, #14845:
URL: https://github.com/apache/lucene/issues/14845
### Description
@benwtrent 's PR at #14836 introduced an estimator for the number of visited
nodes based on k and the number of vectors called `expectedVisitedNodes(int k,
int size)`.
Currently, filtered vector search works roughly this way:
1. compute a per-leaf k value based on the global k value
2. if (filterCost <= perLeafK) do an exact search
3. else run an approximate search, stop after visiting filterCost+1 nodes
4. if the approximate search stopped early (and is thus incomplete), run
an exact search
Could/should we update step 2 above to do an exact search if (filterCost <=
expectedVisitedNodes(perLeafK, size))? This should help save approximate
searches that are almost certainly going to early terminate anyway?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
