diff options
author | Liang-Chi Hsieh <viirya@gmail.com> | 2015-02-02 19:34:25 -0800 |
---|---|---|
committer | Xiangrui Meng <meng@databricks.com> | 2015-02-02 19:34:25 -0800 |
commit | 1bcd46574e442e20f55709d70573f271ce44e5b9 (patch) | |
tree | d54d597053d9aab0191dca30ad2edfa34b402f45 | |
parent | 0561c4544967fb853419f32e014fac9b8879b0db (diff) | |
download | spark-1bcd46574e442e20f55709d70573f271ce44e5b9.tar.gz spark-1bcd46574e442e20f55709d70573f271ce44e5b9.tar.bz2 spark-1bcd46574e442e20f55709d70573f271ce44e5b9.zip |
[SPARK-5512][Mllib] Run the PIC algorithm with initial vector suggected by the PIC paper
As suggested by the paper of Power Iteration Clustering, it is useful to set the initial vector v0 as the degree vector d. This pr tries to add a running method for that.
Author: Liang-Chi Hsieh <viirya@gmail.com>
Closes #4301 from viirya/pic_degreevector and squashes the following commits:
7db28fb [Liang-Chi Hsieh] Refactor it to address comments.
19cf94e [Liang-Chi Hsieh] Add an option to select initialization method.
ec88567 [Liang-Chi Hsieh] Run the PIC algorithm with degree vector d as suggected by the PIC paper.
2 files changed, 47 insertions, 4 deletions
diff --git a/mllib/src/main/scala/org/apache/spark/mllib/clustering/PowerIterationClustering.scala b/mllib/src/main/scala/org/apache/spark/mllib/clustering/PowerIterationClustering.scala index fcb9a3643c..9b5c155b0a 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/clustering/PowerIterationClustering.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/clustering/PowerIterationClustering.scala @@ -43,15 +43,19 @@ class PowerIterationClusteringModel( * * @param k Number of clusters. * @param maxIterations Maximum number of iterations of the PIC algorithm. + * @param initMode Initialization mode. */ class PowerIterationClustering private[clustering] ( private var k: Int, - private var maxIterations: Int) extends Serializable { + private var maxIterations: Int, + private var initMode: String) extends Serializable { import org.apache.spark.mllib.clustering.PowerIterationClustering._ - /** Constructs a PIC instance with default parameters: {k: 2, maxIterations: 100}. */ - def this() = this(k = 2, maxIterations = 100) + /** Constructs a PIC instance with default parameters: {k: 2, maxIterations: 100, + * initMode: "random"}. + */ + def this() = this(k = 2, maxIterations = 100, initMode = "random") /** * Set the number of clusters. @@ -70,6 +74,18 @@ class PowerIterationClustering private[clustering] ( } /** + * Set the initialization mode. This can be either "random" to use a random vector + * as vertex properties, or "degree" to use normalized sum similarities. Default: random. + */ + def setInitializationMode(mode: String): this.type = { + this.initMode = mode match { + case "random" | "degree" => mode + case _ => throw new IllegalArgumentException("Invalid initialization mode: " + mode) + } + this + } + + /** * Run the PIC algorithm. * * @param similarities an RDD of (i, j, s_ij_) tuples representing the affinity matrix, which is @@ -82,7 +98,10 @@ class PowerIterationClustering private[clustering] ( */ def run(similarities: RDD[(Long, Long, Double)]): PowerIterationClusteringModel = { val w = normalize(similarities) - val w0 = randomInit(w) + val w0 = initMode match { + case "random" => randomInit(w) + case "degree" => initDegreeVector(w) + } pic(w0) } @@ -149,6 +168,20 @@ private[clustering] object PowerIterationClustering extends Logging { } /** + * Generates the degree vector as the vertex properties (v0) to start power iteration. + * It is not exactly the node degrees but just the normalized sum similarities. Call it + * as degree vector because it is used in the PIC paper. + * + * @param g a graph representing the normalized affinity matrix (W) + * @return a graph with edges representing W and vertices representing the degree vector + */ + def initDegreeVector(g: Graph[Double, Double]): Graph[Double, Double] = { + val sum = g.vertices.values.sum() + val v0 = g.vertices.mapValues(_ / sum) + GraphImpl.fromExistingRDDs(VertexRDD(v0), g.edges) + } + + /** * Runs power iteration. * @param g input graph with edges representing the normalized affinity matrix (W) and vertices * representing the initial vector of the power iterations. diff --git a/mllib/src/test/scala/org/apache/spark/mllib/clustering/PowerIterationClusteringSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/clustering/PowerIterationClusteringSuite.scala index 2bae465d39..03ecd9ca73 100644 --- a/mllib/src/test/scala/org/apache/spark/mllib/clustering/PowerIterationClusteringSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/mllib/clustering/PowerIterationClusteringSuite.scala @@ -55,6 +55,16 @@ class PowerIterationClusteringSuite extends FunSuite with MLlibTestSparkContext predictions(c) += i } assert(predictions.toSet == Set((0 to 3).toSet, (4 to 15).toSet)) + + val model2 = new PowerIterationClustering() + .setK(2) + .setInitializationMode("degree") + .run(sc.parallelize(similarities, 2)) + val predictions2 = Array.fill(2)(mutable.Set.empty[Long]) + model2.assignments.collect().foreach { case (i, c) => + predictions2(c) += i + } + assert(predictions2.toSet == Set((0 to 3).toSet, (4 to 15).toSet)) } test("normalize and powerIter") { |