Github user mengxr commented on a diff in the pull request:
https://github.com/apache/spark/pull/3022#discussion_r22092939
--- Diff:
mllib/src/main/scala/org/apache/spark/mllib/clustering/GaussianMixtureModelEM.scala
---
@@ -0,0 +1,248 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.spark.mllib.clustering
+
+import breeze.linalg.{DenseVector => BreezeVector, DenseMatrix =>
BreezeMatrix}
+import breeze.linalg.Transpose
+
+import org.apache.spark.rdd.RDD
+import org.apache.spark.mllib.linalg.{Matrices, Vector, Vectors}
+import org.apache.spark.mllib.stat.impl.MultivariateGaussian
+
+import scala.collection.mutable.IndexedSeqView
+
+/**
+ * This class performs expectation maximization for multivariate Gaussian
+ * Mixture Models (GMMs). A GMM represents a composite distribution of
+ * independent Gaussian distributions with associated "mixing" weights
+ * specifying each's contribution to the composite.
+ *
+ * Given a set of sample points, this class will maximize the
log-likelihood
+ * for a mixture of k Gaussians, iterating until the log-likelihood
changes by
+ * less than convergenceTol, or until it has reached the max number of
iterations.
+ * While this process is generally guaranteed to converge, it is not
guaranteed
+ * to find a global optimum.
+ *
+ * @param k The number of independent Gaussians in the mixture model
+ * @param convergenceTol The maximum change in log-likelihood at which
convergence
+ * is considered to have occurred.
+ * @param maxIterations The maximum number of iterations to perform
+ */
+class GaussianMixtureModelEM private (
+ private var k: Int,
+ private var convergenceTol: Double,
+ private var maxIterations: Int) extends Serializable {
+
+ // Type aliases for convenience
+ private type DenseDoubleVector = BreezeVector[Double]
+ private type DenseDoubleMatrix = BreezeMatrix[Double]
+ private type VectorArrayView = IndexedSeqView[DenseDoubleVector,
Array[DenseDoubleVector]]
+
+ private type ExpectationSum = (
+ Array[Double], // log-likelihood in index 0
+ Array[Double], // array of weights
+ Array[DenseDoubleVector], // array of means
+ Array[DenseDoubleMatrix]) // array of cov matrices
+
+ // create a zero'd ExpectationSum instance
+ private def zeroExpectationSum(k: Int, d: Int): ExpectationSum = {
+ (Array(0.0),
+ new Array[Double](k),
+ (0 until k).map(_ => BreezeVector.zeros[Double](d)).toArray,
+ (0 until k).map(_ => BreezeMatrix.zeros[Double](d,d)).toArray)
+ }
+
+ // add two ExpectationSum objects (allowed to use modify m1)
+ // (U, U) => U for aggregation
+ private def addExpectationSums(m1: ExpectationSum, m2: ExpectationSum):
ExpectationSum = {
+ m1._1(0) += m2._1(0)
+ var i = 0
+ while (i < m1._2.length) {
+ m1._2(i) += m2._2(i)
+ m1._3(i) += m2._3(i)
+ m1._4(i) += m2._4(i)
+ i = i + 1
+ }
+ m1
+ }
+
+ // compute cluster contributions for each input point
+ // (U, T) => U for aggregation
+ private def computeExpectation(
+ weights: Array[Double],
+ dists: Array[MultivariateGaussian])
+ (sums: ExpectationSum, x: DenseDoubleVector): ExpectationSum = {
+ val k = sums._2.length
+ val p = weights.zip(dists).map { case (weight, dist) => eps + weight *
dist.pdf(x) }
+ val pSum = p.sum
+ sums._1(0) += math.log(pSum)
+ val xxt = x * new Transpose(x)
+ var i = 0
+ while (i < k) {
+ p(i) /= pSum
+ sums._2(i) += p(i)
+ sums._3(i) += x * p(i)
+ sums._4(i) += xxt * p(i)
+ i = i + 1
+ }
+ sums
+ }
+
+ // number of samples per cluster to use when initializing Gaussians
+ private val nSamples = 5
+
+ // an initializing GMM can be provided rather than using the
+ // default random starting point
+ private var initialGmm: Option[GaussianMixtureModel] = None
+
+ /** A default instance, 2 Gaussians, 100 iterations, 0.01 log-likelihood
threshold */
+ def this() = this(2, 0.01, 100)
--- End diff --
Please move the constructor to the beginning.
---
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
