Github user MLnick commented on the issue:
https://github.com/apache/spark/pull/17306
I commented on the linked JIRA also.
In principle I think this can be a useful enhancement and yes the better
efficiency on the caching side is a good benefit. I'd actually been thinking
about a (simpler) version that would allow multiple`Predictors` to be the last
stages of a pipeline, and return the best from those - so evaluating a set of
algorithms at the same time (but keeping the preceding pipeline stages common).
This is a generalization of that idea.
I think it gels with
[SPARK-19071](https://issues.apache.org/jira/browse/SPARK-19071) and the next
phase of that work that would look at smarter caching & re-use of stages in the
pipeline. Also related to the subtask of that ticket - the parallel CV work in
[SPARK-19357](https://issues.apache.org/jira/browse/SPARK-19357).
Since Spark 2.2 code freeze is imminent, I think this will have to be
delayed a little but it would be good to coordinate with #16774 where
appropriate.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
