[
https://issues.apache.org/jira/browse/SPARK-6004?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14337604#comment-14337604
]
Liang-Chi Hsieh commented on SPARK-6004:
----------------------------------------
Why do we want to stop early? I think we want to get the best model not the
model at the training moment first showing decreasing performance on validation
dataset? If you need to tune the model, in practice you would tune the
iteration number, not the validationTol. But it is still okay to have an option
for it, if not a default behavior.
> Pick the best model when training GradientBoostedTrees with validation
> ----------------------------------------------------------------------
>
> Key: SPARK-6004
> URL: https://issues.apache.org/jira/browse/SPARK-6004
> Project: Spark
> Issue Type: Improvement
> Components: MLlib
> Reporter: Liang-Chi Hsieh
> Priority: Minor
>
> Since the validation error does not change monotonically, in practice, it
> should be proper to pick the best model when training GradientBoostedTrees
> with validation instead of stopping it early.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
