david-sitsky commented on issue #13611: URL: https://github.com/apache/lucene/issues/13611#issuecomment-3346197921
@msokolov and @benwtrent - many thanks for your replies. As an experiment, I re-indexed my data with `maxConn` set to 32. I found an instance of a doc to the query which has the 29th highest score, but it does not appear until I set `k` to 1839. In this index there are 2,594 parent docs, so it is a significant percentage of the index population. Also, the parent doc in this instance only has a single nested `text-embeddings` doc. Is this of interest, or is this just part of the approximate part of the algorithm. So I am not sure what this all means. I do appreciate the algorithm is approximate. I am trying to determine if I want (approximately) the top k results, how many items should I fetch in general to try and get very good recall? -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
