[MADLIB-1315] Define the output of madlib_keras_predict

• Type: New Feature
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: v1.16
• Component/s: Deep Learning
• Assignee: Unassigned
• Reporter: Nandish Jayaram
• Created: 03/Apr/19 19:24
• Updated: 23/Apr/19 17:26
• Resolved: 23/Apr/19 17:26

Description

Currently, we only give class label as output in output table. The following output format can be more informative, following the output format in MLP (http://madlib.apache.org/docs/latest/group__grp__nn.html#predict).
Requirements:
1. Update predict interface to have a new optional param named `pred_type`. It takes two values, `response` or `prob`, and default can be `response`.
1. If `pred_type=response`, create one output column for the predicted class label. Following the convention in MLP, this column is to be named `estimated_COL_NAME` where COL_NAME is the name of the column being predicted.
1. If `pred_type=prob`, create one output column for each distinct class value found in training data (this is captured in `class_values` column from model summary table). The new columns should be named `prob_CLASS` where CLASS is the corresponding class value.