Github user sethah commented on a diff in the pull request:
https://github.com/apache/spark/pull/11989#discussion_r57978546
--- Diff:
mllib/src/main/scala/org/apache/spark/mllib/tree/impl/DecisionTreeMetadata.scala
---
@@ -183,11 +183,40 @@ private[spark] object DecisionTreeMetadata extends
Logging {
}
case _ => featureSubsetStrategy
}
+
+ object featureSubsetNumber {
+ def unapply(strategy: String): Option[Int] = try {
+ val number = strategy.toInt
+ if (0 < number) {
+ Some(number)
+ } else {
+ None
+ }
+ } catch {
+ case _ : java.lang.NumberFormatException => None
+ }
+ }
+
+ object featureSubsetFraction {
+ def unapply(strategy: String): Option[Double] = try {
+ val fraction = strategy.toDouble
+ if (0.0 < fraction && fraction <= 1.0) {
+ Some(fraction)
+ } else {
+ None
+ }
+ } catch {
+ case _ : java.lang.NumberFormatException => None
+ }
+ }
+
val numFeaturesPerNode: Int = _featureSubsetStrategy match {
case "all" => numFeatures
case "sqrt" => math.sqrt(numFeatures).ceil.toInt
case "log2" => math.max(1, (math.log(numFeatures) /
math.log(2)).ceil.toInt)
case "onethird" => (numFeatures / 3.0).ceil.toInt
+ case featureSubsetNumber(number) => if (number > numFeatures)
numFeatures else number
+ case featureSubsetFraction(fraction) => (fraction *
numFeatures).ceil.toInt
--- End diff --
Again, I think regex can simplify things a bit here. I haven't tried the
following, but it should work.
```scala
val isIntRegex = "(\\d+)".r
val isFractionRegex = "(0?\\.\\d+|1.0+)".r
val numFeaturesPerNode: Int = _featureSubsetStrategy match {
case "all" => numFeatures
case "sqrt" => math.sqrt(numFeatures).ceil.toInt
case "log2" => math.max(1, (math.log(numFeatures) /
math.log(2)).ceil.toInt)
case "onethird" => (numFeatures / 3.0).ceil.toInt
case isIntRegex(number) => if (number > numFeatures) numFeatures else
number
case isFractionRegex(fraction) => (fraction * numFeatures).ceil.toInt
}
```
Then we could avoid adding the `featureSubset*` objects.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
