Abstract
Sobolev-type error analysis has recently been intensively studied for radial basis function interpolation. Although the results have been very successful, some limitations have been found. First, the spaces of target functions are not large enough for thecase 1≤p≤∞ to be used practically in some applications. Second, error estimates are confined to the case of finitely smooth radial basis functions. Thus, the primary goal of this paper is to provide Sobolev-type Lp-error bounds (1≤p≤∞) to functions in fractional Sobolev spaces for a wide class of radial functions including some infinitely smooth radial functions. Here an infinitely smooth radial function is required to be conditionally positive definite of a certain order m>0. In addition we provide numerical results that illustrate our theoretical error bounds.
Original language | English |
---|---|
Pages (from-to) | 279-293 |
Number of pages | 15 |
Journal | IMA Journal of Numerical Analysis |
Volume | 32 |
Issue number | 1 |
DOIs | |
State | Published - Jan 2012 |
Bibliographical note
Funding Information:The Basic Science Research Program funded by Ministry of Education, Science, and Technology (2009-0076371 to Y.J.L.); Priority Research Centers Program funded by the Ministry of Education, Science and Technology (2009-0093827 to J.Y.) and Basic Science Research Program funded by the Ministry of Education, Science and Technology through the National Research Foundation of Korea (2009-0073036 to J.Y.).
Keywords
- Sobolev space
- approximation order
- interpolation
- radial basis function
- sampling inequality
- scattered data