msokolov commented on PR #796: URL: https://github.com/apache/lucene/pull/796#issuecomment-1118014031
I'm trying not to steal the thunder of the folks who are actually working on this, but at a high level: we were seeing prefiltering being more expensive than postfiltering (over collecting and then filtering) for the same "yield" of top K, but were able to recover the cost, and flip the balance the other way by using the non-matching nodes to traverse the graph up until some threshold. Basically, we play with how we update the lower bound, allowing it to increase based on non-matching (the filter) nodes, until we get close to where we want to be. I guess the intuition is that nodes that are never going to be included in the top K are still useful for traversing the graph. -- 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 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
