[MADLIB-1313] Add 1-hot encoding support for dependent var in fit

• Type: New Feature
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: v1.16
• Component/s: Deep Learning
• Assignee: Nandish Jayaram
• Reporter: Nandish Jayaram
• Created: 28/Mar/19 23:35
• Updated: 06/Jun/19 00:36
• Resolved: 06/Jun/19 00:36

Description

The current fit function for DL assumes dependent variable is not one-hot encoded. But fit uses data obtained after running minibatch_preprocessor_dl that returns a 1-hot encoded array for each dependent var (https://issues.apache.org/jira/browse/MADLIB-1303). 

Fit should be able to work with this 1-hot encoded data to train the model, and also create a column called class_values in the model summary table to map 1-hot index with a class value.

Predict should then be able to use class_values to figure out the label for a given row in the test table.