Compute Symmetrical Uncertainty in parallel

[Alejandro Alcalde]

In this answer Optimizing Flink transformation
(https://stackoverflow.com/questions/52151715/optimizing-flink-transformation/52225586#52225586):

def symmetricalUncertainty(xy: DataSet[(Double, Double)]): Double = {
val su = xy.reduceGroup { in ⇒
    val invec = in.toVector
    val x = invec map (_._2)
    val y = invec map (_._1)

    val mu = mutualInformation(x, y)
    val Hx = entropy(x)
    val Hy = entropy(y)

    2 * mu / (Hx + Hy)
}

su.collect.head

}

I wrote a function to compute Symmetrical Uncertainty with ReduceGroup.
But its being slow on big data sets.

I've read about Combinable GroupReduceFunctions on Flink's
documentation, and I am tring to write a GroupReduceFunction to compute
Symmetrical Uncertainty:

class MyCombinableGroupReducer
  extends GroupReduceFunction[(Double, Double), Double]
  with GroupCombineFunction[(Double, Double), (Double, Double)]{
  override def reduce(
    in: Iterable[(Double, Double)],
    out: Collector[Double]): Unit =
  {
    val x = in map(_._2)
    val y = in map(_._1)

    // collect...
  }

  override def combine(
    in: Iterable[(Double, Double)],
    out: Collector[(Double, Double)]): Unit =
  {
    // ...
  }
}

It is possible to compute this measure in parallel? I do not know what
should I wrote on reduce and combine functions.

Bests
--
elbauldelprogramador.com