spark-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From dbtsai <...@git.apache.org>
Subject [GitHub] spark pull request #17117: [SPARK-10780][ML] Support initial model for KMean...
Date Wed, 08 Mar 2017 21:20:28 GMT
Github user dbtsai commented on a diff in the pull request:

    https://github.com/apache/spark/pull/17117#discussion_r105025280
  
    --- Diff: mllib/src/main/scala/org/apache/spark/ml/clustering/KMeans.scala ---
    @@ -335,17 +369,70 @@ class KMeans @Since("1.5.0") (
         model
       }
     
    +  /**
    +   * Check validity for interactions between parameters.
    +   */
    +  private def assertInitialModelValid(): Unit = {
    +    if ($(initMode) == MLlibKMeans.K_MEANS_INITIAL_MODEL) {
    +      if (isSet(initialModel)) {
    +        val initialModelK = $(initialModel).parentModel.k
    +        if (initialModelK != $(k)) {
    +          throw new IllegalArgumentException("The initial model's cluster count = " +
    +            s"$initialModelK, mismatched with k = $k.")
    +        }
    +      } else {
    +        throw new IllegalArgumentException("Users must set param initialModel if you
choose " +
    +          "'initialModel' as the initialization algorithm.")
    +      }
    +    } else {
    +      if (isSet(initialModel)) {
    +        logWarning(s"Param initialModel will take no effect when initMode is $initMode.")
    +      }
    +    }
    +  }
    +
       @Since("1.5.0")
       override def transformSchema(schema: StructType): StructType = {
    +    assertInitialModelValid()
    --- End diff --
    
    `transformSchema ` is called in `transform` method, and `model.transform` is called in
computing the summary. I think we should fail it earlier instead of checking it in the end.
Also, it's implicit that it's being checked when computing summary. We should explicitly check
it. 
    
    If we have small logic in checking, I'll have those checking code in `fit` method.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

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


Mime
View raw message