Abstract:
The problem of constructing a hyperbolic interpolation spline can be formulated as a differential multipoint boundary value problem. Its discretization yields a linear system with a five-diagonal matrix, which may be ill-conditioned for unequally spaced data. It is shown that this system can be split into diagonally dominant tridiagonal systems, which are solved without computing hyperbolic functions and admit effective parallelization.
Key words:shape-preserving interpolation, differential multipoint boundary value problem, grid method, discrete hyperbolic spline, parallelization of tridiagonal Gaussian elimination.