moe :
The flink documentation has an example for the usage of the Splitter methods in scala which goes like this:
// Create an Array of K TrainTestDataSets
val dataKFolded: Array[TrainTestDataSet] = Splitter.kFoldSplit(data, 10)
In Java however I am forced to pass to further arguments for TypeInformation
and ClassTag
. I think for the TypeInformation
the TypeInformation.of(TypeofDataset.class)
is the valid input. But I cannot find out what to use for the second parameter.
LppEdd :
Given this Scala signature
def kFoldSplit[T: TypeInformation : ClassTag](
input: DataSet[T],
kFolds: Int,
seed: Long = Utils.RNG.nextLong()
) : Array[TrainTestDataSet[T]]
The Java call would be
Splitter.kFoldSplit(
data,
10,
99,
TypeInformation.of(TypeofDataset.class),
ClassTag$.MODULE$.apply(TypeofDataset.class)
)
ClassTag$
comes from
scala.reflect.ClassTag$
Horrible, but necessary.