1. On the ill-conditioned nature of C∞ RBF strong collocation.
- Author
-
Kansa, E.J. and Holoborodko, P.
- Subjects
- *
RADIAL basis functions , *PARTIAL differential equations , *STOCHASTIC convergence , *POLYNOMIALS , *COMPUTER scientists , *COMPUTER software - Abstract
Continuously differentiable radial basis functions ( C ∞ -RBFs) are the best method to solve numerically higher dimensional partial differential equations (PDEs). Among the reasons are: 1. An n-dimensional problem becomes a one-dimensional radial distance problem, 2. The convergence rate increases with the dimensionality, 3. Such RBFs possess spectral convergence.Finitely supported polynomial methods only converge at polynomial rates. C ∞ -RBFs have global support; the systems of equations may become computationally singular if the condition number exceeds the inverse machine epsilon, ε M . The solution to computational singularity is to decrease the effective ε M by either hardware or software methods. Computer scientists developed rapidly executable multi-precision packages. [ABSTRACT FROM AUTHOR]
- Published
- 2017
- Full Text
- View/download PDF