Github user jkbradley commented on the pull request:
https://github.com/apache/spark/pull/7517#issuecomment-124605394
About abstractions: I'd recommend we just don't use the Regressor
abstraction. I've come to believe that the class hierarchy is a useful
developer tool (for other algorithms), but that it should not be treated as a
standard set of abstractions which need to be used by all algorithms. For the
standard abstractions, we should define some traits (in the next release?).
About linking logreg to isotonic regression: The best option I have thought
of is: Allow isotonic regression to accept Vectors of length 1. Users should
use VectorSlicer to create a new column using value index 1 from logreg's
probabilityCol. This should be documented in isotonic regression.
---
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]
