Résumé:
Radial basis functions have proved very useful in computer graphicx and neutral networks and
are growing in popularity for solving partial differential equations. A small shape parameter and
a small fill distance are both desirable for accuracy, but both cause ill conditioned problems. The
accuracy of RBFs meshless greatly depends on the user defined radial basis centers and the shape
parameter. The researchers are confirmed that even when circumventing the ill conditioning
of the system matrix there usually is a value of the shape parameter which results in optimal
approximation errors. So it is necessary to find a strategy between the good accuracy and the
well posed interpolation problem and therefore looks for a good balance between accuracy and
stability.
