Split out tests

derrickburns · Jan 18, 2024 · 894d15c · 894d15c
1 parent 4010048
commit 894d15c
Showing 1 changed file with 35 additions and 20 deletions.
diff --git a/src/test/scala/com/massivedatascience/clusterer/KMeansSuite.scala b/src/test/scala/com/massivedatascience/clusterer/KMeansSuite.scala
@@ -38,39 +38,65 @@ class KMeansSuite extends AnyFunSuite with LocalClusterSparkContext {
 
   import com.massivedatascience.clusterer.KMeansSelector._
 
-    test("coverage") {
-
+  test("use two runs") {
     val seed = 0
     val r = new Random(seed)
     val data = sc.parallelize[Vector](Array.fill(1000)(Vectors.dense(Array.fill(20)(r.nextDouble()))))
 
-    KMeans.train(data, k = 20, maxIterations = 10)
-
     KMeans.train(data, k = 20, maxIterations = 10, runs = 2)
+  }
 
-    KMeans.train(data, k = 20, maxIterations = 10, runs = 1, mode = KMeansSelector.RANDOM)
+   test("use change tracking clustering algorithm") {
 
-    KMeans.train(data, k = 20, maxIterations = 10, runs = 1, distanceFunctionNames = Seq(BregmanPointOps.EUCLIDEAN))
+    val seed = 0
+    val r = new Random(seed)
+    val data = sc.parallelize[Vector](Array.fill(1000)(Vectors.dense(Array.fill(20)(r.nextDouble()))))
 
     KMeans.train(data, k = 20, maxIterations = 10, runs = 1,
       distanceFunctionNames = Seq(BregmanPointOps.EUCLIDEAN),
       clustererName = MultiKMeansClusterer.CHANGE_TRACKING)
+  }
+  test("use reseed clustering algorithm") {
+    val seed = 0
+    val r = new Random(seed)
+    val data = sc.parallelize[Vector](Array.fill(1000)(Vectors.dense(Array.fill(20)(r.nextDouble()))))
 
     KMeans.train(data, k = 50, maxIterations = 10, runs = 1,
       distanceFunctionNames = Seq(BregmanPointOps.EUCLIDEAN),
       clustererName = MultiKMeansClusterer.RESEED)
+  }
+
+  test ("cluster time series data using Haar transform") {  
+    val seed = 0
+    val r = new Random(seed)
+    val data = sc.parallelize[Vector](Array.fill(1000)(Vectors.dense(Array.fill(20)(r.nextDouble()))))
 
     KMeans.timeSeriesTrain(
       new RunConfig(20, 1, 0, 10),
       data.map(WeightedVector.apply),
       KMeansSelector(KMeansSelector.K_MEANS_PARALLEL),
       BregmanPointOps(BregmanPointOps.EUCLIDEAN),
       MultiKMeansClusterer(MultiKMeansClusterer.COLUMN_TRACKING),
-      Embedding(HAAR_EMBEDDING))
+      Embedding(HAAR_EMBEDDING))  
+  }
 
-    KMeans.train(data, k = 20, maxIterations = 10, runs = 1, clustererName = MultiKMeansClusterer.MINI_BATCH_10)
+  test("use random initializer") {
+    val seed = 0
+    val r = new Random(seed)
+    val data = sc.parallelize[Vector](Array.fill(1000)(Vectors.dense(Array.fill(20)(r.nextDouble()))))
+
+    KMeans.train(data, k = 20, maxIterations = 10, runs = 1, mode = KMeansSelector.RANDOM)
   }
 
+  test("use maximum of 10 iterations") {
+    val seed = 0
+    val r = new Random(seed)
+    val data = sc.parallelize[Vector](Array.fill(1000)(Vectors.dense(Array.fill(20)(r.nextDouble()))))
+
+    KMeans.train(data, k = 20, maxIterations = 10)
+  }
+
+
   test("sparse vector iterator") {
 
     import com.massivedatascience.linalg._
@@ -179,23 +205,12 @@ class KMeansSuite extends AnyFunSuite with LocalClusterSparkContext {
     assert(!iter.hasNext)
   }
 
-
-
-  test("iteratively train") {
+  test("use iterative training") {
     val seed = 0
     val r = new Random(seed)
 
     val data = sc.parallelize[Vector](Array.fill(1000)(Vectors.dense(Array.fill(20)(r.nextDouble()))))
 
-    /*
-
-    runConfig: RunConfig,
-    pointOps: Seq[BregmanPointOps],
-    dataSets: Seq[RDD[BregmanPoint]],
-    initializer: KMeansSelector,
-    clusterer: MultiKMeansClusterer)
-     */
-
     val ops = BregmanPointOps(BregmanPointOps.EUCLIDEAN)
     val cached = data.map(WeightedVector.apply).map(ops.toPoint).cache()
     KMeans.iterativelyTrain(