Github user srowen commented on the issue:
https://github.com/apache/spark/pull/12896
For scoring, returning "I don't know" could be reasonable in order to let
some other logic take over. I could see that. It would be important to do that,
but maybe a separable concern. For evaluation, hm, if the goal is to ignore
cases where the result will reasonably be "I don't know" then what about
constructing the evaluation differently, so that the test set doesn't have
users that the train set doesn't for example? that's more sound, because then
any 'bad' NaNs still show up. Is it that this is hard to implement in the
current structure?
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
