In machine learning math I guess you can make your own dataset from functions, n=np.arange(100),data=[n,np.cumsum(n),n**2] then do a transformation data[:,1] = model(data[:,0],data[:,2]). Then a limited math model() of some sort should show the equation?