[jira] [Assigned] (MADLIB-1452) Add top n to evalute()

["Frank McQuillan (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/MADLIB-1452?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Frank McQuillan reassigned MADLIB-1452:
---------------------------------------

    Assignee: Orhan Kislal

> Add top n to evalute()
> ----------------------
>
>                 Key: MADLIB-1452
>                 URL: https://issues.apache.org/jira/browse/MADLIB-1452
>             Project: Apache MADlib
>          Issue Type: Improvement
>          Components: Deep Learning
>            Reporter: Frank McQuillan
>            Assignee: Orhan Kislal
>            Priority: Major
>             Fix For: v1.18.0
>
>
> Applies to 
> https://madlib.apache.org/docs/latest/group__grp__keras.html#keras_evaluate
> &
> https://madlib.apache.org/docs/latest/group__grp__keras__run__model__selection.html#keras_evaluate
> Add a new parameter to the evaluate interface:
> {code}
> madlib_keras_evaluate(
>     model_table,
>     test_table,
>     output_table,
>     use_gpus,
>     mst_key,
>     top_n                 -- new parameter
>     )
> {code}
> {code}
>  top_n (optional)
> INTEGER[], default {1}.  Array of top values to compute accuracy percentages using the metric from the training set.  E.g., {1, 5, 10} means compute the top-1, top-5 and top-10 classification accuracies.
> {code}
> Add 2 new columns to the right side of the output table:
> {code}
> output_table
> TEXT. Name of table that validation output will be written to. Table contains:
> loss	                                    Loss value on evaluation dataset.
> metric	                            Metric value on evaluation dataset, where 'metrics_type' below identifies the type of metric.
> metrics_type	                    Type of metric used that was used in the training step.
> top_n_accuracy                Array of percentage accuracies as per metric _type
> top_n                                  Array defining the top n values used.
> {code}



--
This message was sent by Atlassian Jira
(v8.3.4#803005)