[
https://issues.apache.org/jira/browse/SPARK-6125?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sean Owen resolved SPARK-6125.
------------------------------
Resolution: Not a Problem
(Despite there being a "Question" JIRA type inherited from the Apache defaults,
this would probably best start as a question on user@.)
Yes you can effectively plug in your own regularization by implementing an
{{Updater}} and plugging it in to a {{GradientDescent}} that you instantiate
and set as the {{Optimizer}} for your algorithm.
If that's not what you need maybe we can continue on the mailing list.
> Custom Loss function
> --------------------
>
> Key: SPARK-6125
> URL: https://issues.apache.org/jira/browse/SPARK-6125
> Project: Spark
> Issue Type: Question
> Components: MLlib
> Affects Versions: 1.2.1
> Reporter: Joanne Shin
> Labels: features, newbie
>
> Currently, there are only a small selection of loss functions available for
> the SGD solver in MLlib. Is there anyway to implement an L1 AND L2
> regularizer or a way to use a custom loss function?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
