Github user srowen commented on the issue: https://github.com/apache/spark/pull/12896 Hm, let me go through the logic with you one more time here. Isn't it better in theory to fix the model to not return NaN, but rather return _some_ default answer, even if it's "0" or equivalent? this is at least no worse for scoring, and fixes the evaluation problem. New users and items are reasonable conditions for this model, not an error case. The current behavior isn't that helpful, so I'm not sure leaving it as a choice is doing anybody a favor. My concern with the "drop" mode is that it is not penalizing any case where the model can't make an answer.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org