Add warnings about using the GLM alpha array internal grid

[exalate-issue-sync]

We allow the user to do an internal grid search for GLM alpha by passing an array instead of a float.

When calling GLM with an array of alpha (what we do in AutoML), a cold start is used at each new alpha value. However, with the same alpha value, lambda-search will have warm start (making use of the previous coefficient). You can actually pass an array of alphas directly to alpha in GLM, however, this is not documented.

Add the warning if user passes alpha to H2O Grid w/ GLM to warn them this is a less efficient way to search alpha

{code:r}> g <- h2o.glm(y = "Species", training_frame = as.h2o(iris), lambda_search = TRUE, alpha = c(0.0, 0.5)){code}

The documentation-only version of this ticket is here: [https://h2oai.atlassian.net/browse/PUBDEV-8171|https://h2oai.atlassian.net/browse/PUBDEV-8171|smart-link]

[h2o-ops]

JIRA Issue Details

Jira Issue: PUBDEV-8150
Assignee: Wendy
Reporter: Erin LeDell
State: Resolved
Fix Version: 3.32.1.4
Attachments: N/A
Development PRs: Available

[h2o-ops]

Linked PRs from JIRA

#5511