msokolov commented on issue #12627: URL: https://github.com/apache/lucene/issues/12627#issuecomment-1812647124
My memory of the way this diversity criterion has evolved is kind of hazy, but I believe in the very first implementation we would not impose any diversity check until the neighbor array was full? It seems as if that would have tended to preserve more links, but then we (I think?) decided that wasn't the approach advocated by the academic papers we were emulating, and implemented the diversity check to every added neighbor. Now we have this checked/unchecked distinction. I seem to remember that the checked ones have had a diversity check applied -- so we know that as a group, they are diverse (none of them is closer to any of the others than it is to the target node). So this would seem to support the idea that we can add unchecked nodes and then only check when the list gets full?? So, what are we doing?! Sorry, I've lost track. -- 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
