msokolov commented on PR #14226: URL: https://github.com/apache/lucene/pull/14226#issuecomment-2807641552
> I'm also in favor of making this the default behavior. Maybe we can let the lambda parameter to be tuned with a reasonable default value (similar to greediness in the current collector) so that expert can tune if they need. From the benchmark we also saw that the tuning range and impact of this optimistic algorithm is larger than the global sharing algorithm, so it may satisfies a wider audiences. Regarding lambda setting, I'm in favor of fixing it as a magic number that replicates something close to the current behavior (or better recall if we can and retain the same latency) and letting users tune further using fanout. I think these are roughly equivalent and I don't think we should be exposing a lot of knobs. Experts who want to can write their own Query. -- 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
