The target function is the (unknown) function which the learning problem 
attempts to approximate.