[
https://issues.apache.org/jira/browse/MADLIB-1338?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Orhan Kislal updated MADLIB-1338:
---------------------------------
Description:
The current {{madlib_keras.fit()}} code reports accuracy as the only metric,
along with loss value. But we could ask for different metrics in compile params
({{mae, binary_accuracy}} etc.), then {{Keras.evaluate()}} would return back
{{loss}} (by default) and {{mean_absolute_error}} or {{binary_accuracy}}
(metrics).
This JIRA requests support to report all of these metrics in the output table.
Other requirements:
Output summary table must have the metrics' labels (instead of just accuracy)
Remove loss/accuracy computation from fit_transition.
was:
The current {{madlib_keras.fit()}} code reports accuracy as the only metric,
along with loss value. But we could ask for multiple metrics in compile params
(for eg., {{metrics=['mae','accuracy']}}), then {{Keras.evaluate()}} would
return back {{loss}} (by default), {{mean_absolute_error}} and {{accuracy}}
(metrics).
This JIRA requests support to report all of these metrics in the output table.
Other requirements:
1. Output summary table must have a 2-D array to report {{metrics}}. The inner
dimension corresponds to all metrics values for the iteration at which it is
computed.
1. Output summary table must have the metrics' labels (eg.,
[mean_absolute_error, accuracy])
Summary: DL: Add support for reporting various metrics in fit/evaluate
(was: DL: Add support for reporting multiple metrics in fit/evaluate)
> DL: Add support for reporting various metrics in fit/evaluate
> -------------------------------------------------------------
>
> Key: MADLIB-1338
> URL: https://issues.apache.org/jira/browse/MADLIB-1338
> Project: Apache MADlib
> Issue Type: New Feature
> Components: Deep Learning
> Reporter: Nandish Jayaram
> Priority: Major
> Fix For: v1.16
>
>
> The current {{madlib_keras.fit()}} code reports accuracy as the only metric,
> along with loss value. But we could ask for different metrics in compile
> params ({{mae, binary_accuracy}} etc.), then {{Keras.evaluate()}} would
> return back {{loss}} (by default) and {{mean_absolute_error}} or
> {{binary_accuracy}} (metrics).
> This JIRA requests support to report all of these metrics in the output table.
> Other requirements:
> Output summary table must have the metrics' labels (instead of just accuracy)
> Remove loss/accuracy computation from fit_transition.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
