[jira] [Updated] (SPARK-6867) Dropout regularization

["Sean Owen (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/SPARK-6867?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sean Owen updated SPARK-6867:
-----------------------------
    Target Version/s:   (was: 1.4.0)

> Dropout regularization
> ----------------------
>
>                 Key: SPARK-6867
>                 URL: https://issues.apache.org/jira/browse/SPARK-6867
>             Project: Spark
>          Issue Type: New Feature
>          Components: MLlib
>            Reporter: Rakesh Chalasani
>            Priority: Minor
>
> Linear models is MLLIB so far support no regularization, L1 and L2. Another more recently popularized method for regularization is dropout [http://www.cs.toronto.edu/~rsalakhu/papers/srivastava14a.pdf]. The dropout regularization basically randomly omit some of the input features at each iteration. 
> Though this approach is particularly used in training deep networks, they could also be very useful on a linear models as if promotes adaptive regularization. This approach is particularly useful in NLP [http://papers.nips.cc/paper/4882-dropout-training-as-adaptive-regularization.pdf] and, because of its simplicity can be easily adopted for streaming linear models as well.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org