Github user MLnick commented on the issue:
https://github.com/apache/spark/pull/17090
Fitting into the CV / evaluator is actually fairly straightforward. It's
just that the semantics of `transform` for top-k recommendation must fit into
whatever we decide on for `RankingEvaluator`, so they are closely linked. (In
other words, they must be compatible). Once the semantics (basically output
schema for `transform`) are decided it's quite simple.
It was discussed on the JIRA
[here](https://issues.apache.org/jira/browse/SPARK-13857?focusedCommentId=15236796&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15236796)
and
[here](https://issues.apache.org/jira/browse/SPARK-13857?focusedCommentId=15822021&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15822021).
I haven't had a chance to refine it yet but I have a view on the best
approach now (basically to fit in with the design of
[SPARK-14409](https://issues.apache.org/jira/browse/SPARK-14409) and in
particular the basic version of #16618). I think that design / schema is more
"DataFrame-like".
In any event - I'm not against having the convenience methods for
recommend-all here. I support it. Ultimately the `transform` approach is mostly
for fitting into Pipelines & cross-validation. `transform` could call into
these convenience methods (though it will need a DataFrame-based input version).
---
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]
