srowen commented on a change in pull request #21632: [SPARK-19591][ML][MLlib]
Add sample weights to decision trees
URL: https://github.com/apache/spark/pull/21632#discussion_r243594110
##########
File path: mllib/src/main/scala/org/apache/spark/ml/tree/impl/BaggedPoint.scala
##########
@@ -33,13 +33,20 @@ import org.apache.spark.util.random.XORShiftRandom
* this datum has 1 copy, 0 copies, and 4 copies in the 3 subsamples,
respectively.
*
* @param datum Data instance
- * @param subsampleWeights Weight of this instance in each subsampled dataset.
- *
- * TODO: This does not currently support (Double) weighted instances. Once
MLlib has weighted
- * dataset support, update. (We store subsampleWeights as Double for
this future extension.)
+ * @param subsampleCounts Number of samples of this instance in each
subsampled dataset.
+ * @param sampleWeight The weight of this instance.
*/
-private[spark] class BaggedPoint[Datum](val datum: Datum, val
subsampleWeights: Array[Double])
- extends Serializable
+private[spark] class BaggedPoint[Datum](
+ val datum: Datum,
+ val subsampleCounts: Array[Int],
+ val sampleWeight: Double) extends Serializable {
+
+ /**
+ * Subsample counts weighted by the sample weight.
+ */
+ def weightedCounts: Array[Double] = subsampleCounts.map(_ * sampleWeight)
Review comment:
You can just store subsampleCounts * sampleWeights as a field, and then
these two constructor params won't become fields. No need to compute it each
time. On the other hand I don't mind this change as it reduces memory usage --
ints are smaller than doubles.
On the other other hand, if you often call this method with one sample, it
could be more efficient, to even provide a second constructor taking an Int
rather than array. Then you do need to store the array of doubles internally.
I don't feel strongly about it
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
