[
https://issues.apache.org/jira/browse/MADLIB-1422?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17069062#comment-17069062
]
Frank McQuillan commented on MADLIB-1422:
-----------------------------------------
Thank you for the JIRA.
The focus so far for DL on MADlib has been image classification. But we need
to look at regression in more detail; I feel like we might be able to support
this today, but need to investigate some more exactly how to do that. Will do
so and post my findings to this JIRA.
--------------
Here are the user docs for the preprocessor
https://github.com/apache/madlib/blob/master/src/ports/postgres/modules/deep_learning/input_data_preprocessor.sql_in
dependent_varname
TEXT. Name of the dependent variable column.
@note The mini-batch preprocessor automatically 1-hot encodes
dependent variables of all types. The exception is numeric array types
(integer and float), where we assume these are already 1-hot encoded,
so these will just be passed through as is.
> Support regression in MADlib keras
> ----------------------------------
>
> Key: MADLIB-1422
> URL: https://issues.apache.org/jira/browse/MADLIB-1422
> Project: Apache MADlib
> Issue Type: New Feature
> Reporter: Hubert Zhang
> Priority: Major
>
> Currently, we can only do classification on MADlib with keras.
> Users have to map interval to category and then simulate the regression by
> classification. Better to support regression as well.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
