[
https://issues.apache.org/jira/browse/FLINK-1994?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15090246#comment-15090246
]
ASF GitHub Bot commented on FLINK-1994:
---------------------------------------
Github user rawkintrevo commented on the pull request:
https://github.com/apache/flink/pull/1397#issuecomment-170162685
on your second point- I agree that Enum would be better for human readable
variable setting, but if one was doing a grid parameter search it would be
easier to search over a range of integers than having to establish an array of
specifically named variables. Not significantly harder, but still slightly more
tedious.
Further, since I'm just learning scala- limited research/understanding
leads me to believe 1) scala doesn't have an enum that works just like Java,
and 2) changing the cases to more informative string names would achieve the
desired goal?
In the mean time working on documentation and reverting the indentation
(which I thought I had already done), and thanks for bearing with me as I am
still learning.
> 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)
