Parallel training using different response columns

[exalate-issue-sync[bot]]

Add support for parallel training (train several models at once of same or different algo) using the same training data, but with different response columns.

[exalate-issue-sync[bot]]

Mark Landry commented: Part of a series related to alternatives to the method of distribution (algorithm level) PUBDEV-1226 Build trees in parallel on all the data for DRF PUBDEV-1979 Train N-fold CV models in parallel PUBDEV-2509 Parallel training using different response columns PUBDEV-2510 Run grid search models in parallel

[exalate-issue-sync[bot]]

Michal Malohlava commented: The response parameter is configured as gridable on the backend (design decision on backend is not to limit gridable parameters). So it should work.

But i expect that R/Python clients expect single response.

@eric do we have R/Python test for this scenario? Or all the tests have fixed single response?

[DinukaH2O]

JIRA Issue Migration Info

Jira Issue: PUBDEV-2509 Assignee: New H2O Bugs Reporter: Erin LeDell State: Open Fix Version: N/A Attachments: N/A Development PRs: N/A