Often in binary classification we are interested in the differences in the output of our current

classifier, g$,$ and an unknown function f that we are trying to learn.