jmazanec15 commented on issue #12440: URL: https://github.com/apache/lucene/issues/12440#issuecomment-1703111158
> Another idea I have is a bit wild: what if we do less merge? > > For example, if we have segment 1,2,3,4 wants to merge and form a new segment, can we just leave the HNSW graphs as-is, or we only merge the smaller ones and leave the big ones as-is. And when we do searching, we can do a greedy search (always go to the closest node, whichever graph), or we can just let user choose to use multithreading to exchange for the latency? Would this potentially lead towards a bias of exploring small graphs? For instance, when searching, the small graph may converge to its nearest neighbors before the larger graph and then prevent the intermediate nodes in the larger graph from ever getting visited? -- 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
