If it is desired the built function belong to a particular class or hold certain properties, additional constraints can be added when fitting. In particular, we can ensure
monotonicity holds by expressing linear conditions on the coefficients c j1 j2... jn. The
fitting of this function to data involves sparse matrices, their size increasing with the
number of basis functions in respect to each variable and exponentially with n. We
give an example of this fitting process in the Example 22.10.