This library of C functions computes approximations of sums of the form
based on the recently developed fast Fourier transform at
nonequispaced knots.
Here are special kernels, e.g.,
and in its multivariate generalization for RBFs .
Our algorithm can be modified for other kernels frequently
used in the approximation by RBFs, e.g., the Gaussian
or the (inverse) multiquadric
.
New kernels are easily incorporated
by defining an appropriate C-function (see kernels.c for some examples).
fastsum_test.c is an example for the usage of the library. The MATLAB script
file fastsum_test.m calls the MATLAB function fastsum.m, which is a simple
example for the usage in MATLAB.
In summary it requires
arithmetic operations.
The algorithms are implemented by Markus Fenn in ./applications/fastsum.
Related paper are
Potts, D. and Steidl, G. Fast summation at nonequispaced knots by NFFTs.
SIAM J. on Sci. Comput. 24, 2013-2037. (full paper ps, pdf), 2004
Potts, D., Steidl, G. and Nieslony, A. Fast convolution with radial kernels at nonequispaced knots.
Numer. Math. 98, 329-351. (full paper ps, pdf), 2004
Fenn, M. and Steidl, G. Fast NFFT based summation of radial functions. Sampling Theory in Signal and Image Processing 3/1, 1-28 (2004)
Fenn, M. and Potts, D. Fast summation based on fast trigonometric transforms at nonequispaced nodes.
Numer. Linear Algebra Appl. 12, 161-169.
(full paper ps, pdf), 2005
Pöplau, G., Potts, D. and van Rienen, U. Calculation of 3D Space-Charge Fields of Bunches of Charged Particles by Fast Summation.
in: Proceedings of SCEE 2004 (5th International Workshop on Scientific Computing in Electrical Engineering), Springer-Verlag
(full paper ps,pdf
), 2005