Github user dbtsai commented on the pull request:
https://github.com/apache/spark/pull/4259#issuecomment-72985804
I just rebased and addressed the api changes.
I was thinking about that when we do standardization on the feature, we're
actually rescaling on the data which is basically rescaling on the covariant
position of the equation. However, we can achieve the same scaling on rescaling
on the graidentSum which is contravariant position. Thus, we don't need to
apply the scaler on the data which will be much cheaper. This also works for
the logistic regression as well. For intercept in LOR, we can deal with it in
gradient function instead of using applyBias, so combining both of those two
techniques, we don't have to create new dataset.
---
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]
