[
https://issues.apache.org/jira/browse/FLINK-1994?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15090670#comment-15090670
]
ASF GitHub Bot commented on FLINK-1994:
---------------------------------------
Github user rawkintrevo commented on the pull request:
https://github.com/apache/flink/pull/1397#issuecomment-170257023
still refactoring to enumeration, but wanted to toss this up in case anyone
is watching- I added a decay parameter which significantly generalizes the Xu
and Inverse Scaling Methods. Also I added some docs, but am open to
suggestions, for example- once I figure out how to set the method with
enumeration, I'll include a note?
> Add different gain calculation schemes to SGD
> ---------------------------------------------
>
> Key: FLINK-1994
> URL: https://issues.apache.org/jira/browse/FLINK-1994
> Project: Flink
> Issue Type: Improvement
> Components: Machine Learning Library
> Reporter: Till Rohrmann
> Assignee: Trevor Grant
> Priority: Minor
> Labels: ML, Starter
>
> The current SGD implementation uses as gain for the weight updates the
> formula {{stepsize/sqrt(iterationNumber)}}. It would be good to make the gain
> calculation configurable and to provide different strategies for that. For
> example:
> * stepsize/(1 + iterationNumber)
> * stepsize*(1 + regularization * stepsize * iterationNumber)^(-3/4)
> See also how to properly select the gains [1].
> Resources:
> [1] http://arxiv.org/pdf/1107.2490.pdf
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
