[
https://issues.apache.org/jira/browse/MADLIB-950?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15705956#comment-15705956
]
Frank McQuillan commented on MADLIB-950:
----------------------------------------
Ivan, thanks for the question.
Grouping is one of the great thing about SQL based machine learning since it
allows one to build multiple models at the same time. For example, if I had
data for 50 different investment products and I wanted a different predictive
model for each product, I could GROUP BY investment product in MADlib to
achieve that.
Most MADlib algos support grouping but elastic net did not up to this date, for
some reason. PCA is another example that we are fixing in v1.10.
https://issues.apache.org/jira/browse/MADLIB-947
Now, one could achieve grouping by putting a FOR loop around things, but that
is less performant since it runs in series. By using the MADlib group
iteration component, we can use the query processor to run multiple models in
parallel.
> Add grouping support to elastic net
> -----------------------------------
>
> Key: MADLIB-950
> URL: https://issues.apache.org/jira/browse/MADLIB-950
> Project: Apache MADlib
> Issue Type: New Feature
> Components: Module: Regularized Regression
> Reporter: Frank McQuillan
> Assignee: Nandish Jayaram
> Fix For: v1.10
>
>
> http://doc.madlib.net/latest/group__grp__elasticnet.html
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
