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

    https://github.com/apache/spark/pull/11119#discussion_r83509799
  
    --- Diff: 
mllib/src/test/scala/org/apache/spark/ml/clustering/KMeansSuite.scala ---
    @@ -137,18 +143,64 @@ class KMeansSuite extends SparkFunSuite with 
MLlibTestSparkContext with DefaultR
           assert(model.clusterCenters === model2.clusterCenters)
         }
         val kmeans = new KMeans()
    -    testEstimatorAndModelReadWrite(kmeans, dataset, 
KMeansSuite.allParamSettings, checkModelData)
    +    testEstimatorAndModelReadWrite(kmeans, dataset, 
KMeansSuite.allParamSettings, checkModelData,
    +      Map("initialModel" -> (checkModelData _).asInstanceOf[(Any, Any) => 
Unit]))
    +  }
    +
    +  test("Initialize using a trained model") {
    +    val kmeans = new KMeans().setK(k).setSeed(1).setMaxIter(1)
    +    val oneIterModel = kmeans.fit(dataset)
    +    val twoIterModel = kmeans.copy(ParamMap(ParamPair(kmeans.maxIter, 
2))).fit(dataset)
    +    val oneMoreIterModel = 
kmeans.setInitialModel(oneIterModel).fit(dataset)
    +
    +    twoIterModel.clusterCenters.zip(oneMoreIterModel.clusterCenters)
    +      .foreach { case (center1, center2) => assert(center1 ~== center2 
absTol 1E-8) }
    +  }
    +
    +  test("Initialize using a model with wrong dimension of cluster centers") 
{
    +    val kmeans = new KMeans().setK(k).setSeed(1).setMaxIter(1)
    +
    +    val wrongDimModel = KMeansSuite.generateRandomKMeansModel(4, k)
    +    val wrongDimModelThrown = intercept[IllegalArgumentException] {
    +      kmeans.setInitialModel(wrongDimModel).fit(dataset)
    +    }
    +    assert(wrongDimModelThrown.getMessage.contains("mismatched dimension"))
    +  }
    +
    +  test("Infer K from an initial model if K is unset") {
    +    val kmeans = new KMeans()
    +    val testNewK = 10
    +    val randomModel = KMeansSuite.generateRandomKMeansModel(dim, testNewK)
    +    assert(kmeans.setInitialModel(randomModel).getK === testNewK)
    +  }
    +
    +  test("Initialize using a model with wrong K if K is set") {
    +    val kmeans = new KMeans().setK(k).setSeed(1).setMaxIter(1)
    +
    +    val wrongKModel = KMeansSuite.generateRandomKMeansModel(3, k + 1)
    +    val wrongKModelThrown = intercept[IllegalArgumentException] {
    +      kmeans.setInitialModel(wrongKModel).fit(dataset)
    --- End diff --
    
    According to discussions below, `setInitialModel` only overwrites param k.
    
    1. If we use `km.setInitialModel(m1)` or `km.setK(k).setInitialModel(m1)`, 
then km with `k=m1.getK`;
    
    2. If we use `km.setInitialModel(m1).setInitialModel(m2)`, then km with `k 
= m2.getK`;
    
    3. If we use `km.setInitialModel(m1).setK(differentK)`, then an exception 
will be thrown out when `km.fit(data)`.
    
    I think the case 3 is a wrong use case, so it's reasonable to throw an 
exception. What do you think?


---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---

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

Reply via email to