[ 
https://issues.apache.org/jira/browse/FLINK-1992?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14553963#comment-14553963
 ] 

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_r30785863
  
    --- Diff: 
flink-staging/flink-ml/src/main/scala/org/apache/flink/ml/optimization/Solver.scala
 ---
    @@ -40,6 +44,32 @@ abstract class Solver extends Serializable with 
WithParameters {
         data: DataSet[LabeledVector],
         initialWeights: Option[DataSet[WeightVector]]): DataSet[WeightVector]
     
    +  /** Creates initial weights vector, creating a DataSet with a 
WeightVector element
    +    *
    +    * @param initialWeights An Option that may contain an initial set of 
weights
    +    * @param data The data for which we optimize the weights
    +    * @return A DataSet containing a single WeightVector element
    +    */
    +  def createInitialWeightsDS(initialWeights: Option[DataSet[WeightVector]],
    +                             data: DataSet[LabeledVector]):  
DataSet[WeightVector] = {
    +    // TODO: Faster way to do this?
    +    val dimensionsDS = data.map(_.vector.size).reduce((a, b) => b)
    +
    +    initialWeights match {
    +      // Ensure provided weight vector is a DenseVector
    +      case Some(wvDS) =>
    +        wvDS.map { wv => {
    +          val denseWeights = wv.weights match {
    +            case dv: DenseVector => dv
    +            case sv: SparseVector => sv.toDenseVector
    +          }
    +          WeightVector(denseWeights, wv.intercept)
    --- End diff --
    
    Not sure what you mean here, is it `WeightVector(denseWeights, 
wv.intercept)` that is off?


> 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)

Reply via email to