[SPARK-8481] [MLLIB] GaussianMixtureModel.predict, GaussianMixtureModel.predictSoft variants for a single vector

This PR adds GaussianMixtureModel.predict & GaussianMixtureModel.predictSoft variants for a single vector which are useful when applying the trained model in environments where spark context is not required (or not desired) and predictions are made for single data points (vectors).
Test case included.

Author: Dariusz Kobylarz <[email protected]>

Closes #6906 from dkobylarz/branch-1.4 and squashes the following commits:

cef1f0a [Dariusz Kobylarz] [SPARK-8481] [MLlib] GaussianMixtureModel predict accepting single vector

Author: dkobylarz

mllib/src/main/scala/org/apache/spark/mllib/clustering/GaussianMixtureModel.scala

@@ -66,6 +66,12 @@ class GaussianMixtureModel(
     responsibilityMatrix.map(r => r.indexOf(r.max))
   }
 
+  /** Maps given point to its cluster index. */
+  def predict(point: Vector): Int = {
+    val r = computeSoftAssignments(point.toBreeze.toDenseVector, gaussians, weights, k)
+    r.indexOf(r.max)
+  }
+
   /** Java-friendly version of [[predict()]] */
   def predict(points: JavaRDD[Vector]): JavaRDD[java.lang.Integer] =
     predict(points.rdd).toJavaRDD().asInstanceOf[JavaRDD[java.lang.Integer]]
@@ -83,6 +89,13 @@ class GaussianMixtureModel(
     }
   }
 
+  /**
+   * Given the input vector, return the membership values to all mixture components.
+   */
+  def predictSoft(point: Vector): Array[Double] = {
+    computeSoftAssignments(point.toBreeze.toDenseVector, gaussians, weights, k)
+  }
+
   /**
    * Compute the partial assignments for each vector
    */

mllib/src/test/scala/org/apache/spark/mllib/clustering/GaussianMixtureSuite.scala

@@ -148,6 +148,16 @@ class GaussianMixtureSuite extends SparkFunSuite with MLlibTestSparkContext {
     }
   }
 
+  test("model prediction, parallel and local") {
+    val data = sc.parallelize(GaussianTestData.data)
+    val gmm = new GaussianMixture().setK(2).setSeed(0).run(data)
+
+    val batchPredictions = gmm.predict(data)
+    batchPredictions.zip(data).collect().foreach { case (batchPred, datum) =>
+      assert(batchPred === gmm.predict(datum))
+    }
+  }
+
   object GaussianTestData {
 
     val data = Array(