I'm downstream stages the labels & features are generally expected to be doubles, so its easier to use as a double.
On Sat, Jan 21, 2017 at 5:32 PM Shiyuan <gshy2...@gmail.com> wrote: > Hi Spark, > StringIndex uses double instead of int for indexing > http://spark.apache.org/docs/latest/ml-features.html#stringindexer. > What's the rationale for using double to index? Would it be more > appropriate to use int to index (which is consistent with other place > like Vector.sparse) > > Shiyuan > > >
