Github user NathanHowell commented on the pull request:
https://github.com/apache/spark/pull/8246#issuecomment-139339009
@jkbradley yeah, I figured out as much. a lot of the internals also assume
dense feature space, which isn't very common with a large number of features.
i tried it with `useNodeIdCache` on and off and it didn't make much of a
difference. checkpointing was also enabled... the dataset fits in memory though
(8TB was allocated to RDD storage), i don't know how much that changes things.
the lower levels of the trees (near the roots) look similar but i haven't
done a thorough comparison of the two.
---
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]
