A fast vector assembler.
A fast vector assembler. The columns given must be ordered such that categorical columns come first
(otherwise spark learners will give categorical attributes to the wrong index).
Does not keep spurious numeric data which can significantly slow down computations when there are
millions of columns.