flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FLINK-1992) Add convergence criterion to SGD optimizer
Date Thu, 21 May 2015 09:00:00 GMT

    [ https://issues.apache.org/jira/browse/FLINK-1992?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14553918#comment-14553918
] 

ASF GitHub Bot commented on FLINK-1992:
---------------------------------------

Github user thvasilo commented on a diff in the pull request:

    https://github.com/apache/flink/pull/692#discussion_r30784142
  
    --- Diff: flink-staging/flink-ml/src/main/scala/org/apache/flink/ml/optimization/GradientDescent.scala
---
    @@ -36,19 +36,20 @@ import org.apache.flink.ml.optimization.Solver._
       * At the moment, the whole partition is used for SGD, making it effectively a batch
gradient
       * descent. Once a sampling operator has been introduced, the algorithm can be optimized
       *
    -  * @param runParameters The parameters to tune the algorithm. Currently these include:
    -  *                      [[Solver.LossFunction]] for the loss function to be used,
    -  *                      [[Solver.RegularizationType]] for the type of regularization,
    -  *                      [[Solver.RegularizationParameter]] for the regularization parameter,
    +  *  The parameters to tune the algorithm are:
    +  *                      [[Solver.LossFunctionParameter]] for the loss function to be
used,
    +  *                      [[Solver.RegularizationTypeParameter]] for the type of regularization,
    +  *                      [[Solver.RegularizationValueParameter]] for the regularization
parameter,
    --- End diff --
    
    I'm doing this to avoid confusion between the LossFunction class and the parameter setting
here. Having the same name I think creates some confusion, especially when auto-completing
stuff in Intellij. However the names are starting to get pretty long.
    
    You think we should remove this?


> Add convergence criterion to SGD optimizer
> ------------------------------------------
>
>                 Key: FLINK-1992
>                 URL: https://issues.apache.org/jira/browse/FLINK-1992
>             Project: Flink
>          Issue Type: Improvement
>          Components: Machine Learning Library
>            Reporter: Till Rohrmann
>            Assignee: Theodore Vasiloudis
>            Priority: Minor
>              Labels: ML
>             Fix For: 0.9
>
>
> Currently, Flink's SGD optimizer runs for a fixed number of iterations. It would be good
to support a dynamic convergence criterion, too.



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

Mime
View raw message