ExpectationSum

logLikelihood：似然函数

weights：每个类的权重

means：每个类的均值

sigmas：每个类的covariance matrix

package org.apache.spark.mllib.clustering

import breeze.linalg.{DenseMatrix => BDM, DenseVector => BDV, Vector => BV}
import org.apache.spark.mllib.linalg.{DenseMatrix, Matrices, Vectors, BLAS}
import org.apache.spark.mllib.stat.distribution.MultivariateGaussian
import org.apache.spark.mllib.util.MLUtils

/**
  * Created by fhqplzj on 16-7-29 at 下午3:29.
  */
class MyExpectationSum(var logLikelihood: Double,
                       val weights: Array[Double],
                       val means: Array[BDV[Double]],
                       val sigmas: Array[BDM[Double]]) extends Serializable {
  def +=(x: MyExpectationSum) = {
    for (i <- weights.indices) {
      weights(i) += x.weights(i)
      means(i) += x.means(i)
      sigmas(i) += x.sigmas(i)
    }
    logLikelihood += x.logLikelihood
    this
  }
}

object MyExpectationSum {
  def zero(k: Int, d: Int) = {
    new MyExpectationSum(0, Array.fill(k)(0), Array.fill(k)(BDV.zeros(d)), Array.fill(k)(BDM.zeros(d, d)))
  }

  def add(weights: Array[Double], dists: Array[MultivariateGaussian])(sums: MyExpectationSum, x: BV[Double]) = {
    val p = weights.zip(dists).map {
      case (weight, dist) =>
        MLUtils.EPSILON + weight * dist.pdf(x)
    }
    val pSum = p.sum
    sums.logLikelihood += math.log(pSum)
    for (i <- p.indices) {
      p(i) /= pSum
      sums.weights(i) += p(i)
      sums.means(i) += p(i) * x
      BLAS.syr(p(i), Vectors.fromBreeze(x), Matrices.fromBreeze(sums.sigmas(i)).asInstanceOf[DenseMatrix])
    }
    sums
  }
}