Github user sethah commented on the issue: https://github.com/apache/spark/pull/19106 I'm confused how this issue was discovered in the first place. Did someone actually train an RF/DT and receive all zero probabilities? If so, shouldn't there be a unit test that recreates that scenario? Anyway, AFAICT the only way this could happen is if the RandomForest was trained on an empty DataFrame which couldn't happen because it would fail before the train method was called anyway.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org