vandermonde_interp_2d, a FORTRAN90 code which finds P(X,Y), a polynomial interpolant to data Z(X,Y) which depends on two independent variables X and Y, by setting up and solving a linear system involving the Vandermonde matrix.
This software is primarily intended as an illustration of the problems that can occur when the interpolation problem is naively formulated using the Vandermonde matrix. Unless the data points are well separated, and the degree of the polynomial is low, the linear system will become very difficult to store and solve accurately, because the monomials used as basis vectors by the Vandermonde approach become indistinguishable.
If the data is available on a product grid, then both the LAGRANGE_INTERP_2D and VANDERMONDE_INTERP_2D libraries will be trying to compute the same interpolating function. However, especially for higher degree polynomials, the Lagrange approach will be superior because it avoids the badly conditioned Vandermonde matrix associated with the usage of monomials as the basis. The Lagrange approach uses as a basis a set of Lagrange basis polynomials l(i,j)(x) which are 1 at node (x(i),y(j)) and zero at the other nodes.
The code needs access to the QR_SOLVE and R8LIB libraries. The test code also needs access to the TEST_INTERP_2D library.
The computer code and data files described and made available on this web page are distributed under the MIT license
vandermonde_interp_2d is available in a C version and a C++ version and a FORTRAN90 version and a MATLAB version.
LAGRANGE_INTERP_2D, a FORTRAN90 code which defines and evaluates the Lagrange polynomial p(x,y) which interpolates a set of data depending on a 2D argument that was evaluated on a product grid, so that p(x(i),y(j)) = z(i,j).
PWL_INTERP_2D, a FORTRAN90 code which evaluates a piecewise linear interpolant to data defined on a regular 2D grid.
QR_SOLVE, a FORTRAN90 code which computes the least squares solution of a linear system A*x=b.
R8LIB, a FORTRAN90 code which contains many utility routines, using double precision real (R8) arithmetic.
RBF_INTERP_2D, a FORTRAN90 code which defines and evaluates radial basis function (RBF) interpolants to 2D data.
SHEPARD_INTERP_2D, a FORTRAN90 code which defines and evaluates Shepard interpolants to 2D data, which are based on inverse distance weighting.
TEST_INTERP_2D, a FORTRAN90 code which defines test problems for interpolation of data z(x,y) of a 2D argument.
TOMS886, a FORTRAN90 code which defines the Padua points for interpolation in a 2D region, including the rectangle, triangle, and ellipse, by Marco Caliari, Stefano de Marchi, Marco Vianello. This is ACM TOMS algorithm 886.
VANDERMONDE_APPROX_2D, a FORTRAN90 code which finds a polynomial approximant p(x,y) to data z(x,y) of a 2D argument by setting up and solving an overdetermined linear system for the polynomial coefficients involving the Vandermonde matrix.
VANDERMONDE_INTERP_1D, a FORTRAN90 code which finds a polynomial interpolant to a function of 1D data by setting up and solving a linear system for the polynomial coefficients, involving the Vandermonde matrix.