[SPARK-18036] Decision Trees do not handle edge cases

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 2.2.0
• Component/s: ML, MLlib
• Assignee: Ilya Matiach
• Reporter: Seth Hendrickson
• Created: 20/Oct/16 22:40
• Updated: 24/Jan/17 18:25
• Resolved: 24/Jan/17 18:25

Description

Decision trees/GBT/RF do not handle edge cases such as constant features or empty features. For example:

val dt = new DecisionTreeRegressor()
val data = Seq(LabeledPoint(1.0, Vectors.dense(Array.empty[Double]))).toDF()
dt.fit(data)

java.lang.UnsupportedOperationException: empty.max
  at scala.collection.TraversableOnce$class.max(TraversableOnce.scala:229)
  at scala.collection.mutable.ArrayOps$ofInt.max(ArrayOps.scala:234)
  at org.apache.spark.ml.tree.impl.DecisionTreeMetadata$.buildMetadata(DecisionTreeMetadata.scala:207)
  at org.apache.spark.ml.tree.impl.RandomForest$.run(RandomForest.scala:105)
  at org.apache.spark.ml.regression.DecisionTreeRegressor.train(DecisionTreeRegressor.scala:93)
  at org.apache.spark.ml.regression.DecisionTreeRegressor.train(DecisionTreeRegressor.scala:46)
  at org.apache.spark.ml.Predictor.fit(Predictor.scala:90)
  ... 52 elided

as well as

val dt = new DecisionTreeRegressor()
val data = Seq(LabeledPoint(1.0, Vectors.dense(0.0, 0.0, 0.0))).toDF()
dt.fit(data)

java.lang.UnsupportedOperationException: empty.maxBy
at scala.collection.TraversableOnce$class.maxBy(TraversableOnce.scala:236)
at scala.collection.SeqViewLike$AbstractTransformed.maxBy(SeqViewLike.scala:37)
at org.apache.spark.ml.tree.impl.RandomForest$.binsToBestSplit(RandomForest.scala:846)

Issue Links

• links to: [Github] Pull Request #16377 (imatiach-msft)