Multiply monotone functions for radial basis function interpolation: extensions and new kernels (Q2182919)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Multiply monotone functions for radial basis function interpolation: extensions and new kernels |
scientific article |
Statements
Multiply monotone functions for radial basis function interpolation: extensions and new kernels (English)
0 references
26 May 2020
0 references
A radial basis function \(\phi\in C([0, \infty))\) is said to be strictly positive definite on \(\mathbb{R}^d\) if and only if for any finite set of different points \(A\subset \mathbb{R}^d,\) the symmetric matrix \([ \phi( || a - b||)] _ {a,b \in A}\) is positive definite. On the other hand, a function \(g\in C^{n-2}(I),\) \(n\) a positive integer number greater than or equal to 2, defined on an interval \(I\) of the real line is called \(n\)-times monotone (or multiply monotone) if and only if \((-1)^j g^{(j)} (t)\geq 0, t\in I.\) Finally, a function \(g:I \rightarrow\mathbb{R}\) is said to be completely monotone on and interval \(I\) of the real line if and only if \(g\in C^{\infty}(I)\) and \((-1)^j g^{(j)}(t) \geq0, t \in I, j\in \mathbb{N}_{0}\). The paper under review focus the attention on the connections of monotonicity properties and the strict positive definiteness of functions on \(\mathbb{R}^d.\) As a first result, the authors prove a constructive method to find a strictly positive definite kernel on the \(2n+1\) dimensional real space as the Fourier cosine transform of a function \(f(x)= g(x^2)\) assuming that \(g\) is a \(n\)-times monotone function on the nonnegative real half-axis and \(f\in L^{1}((0, \infty)) \)(Theorem 2.4). In the same way, in Theorem 2.5 functions which are strictly positive definite in a space whose dimension is limited by an even integer number \(2n+2\) assuming \(g\) is a \(n+\frac{1}{2}\)-times monotone function are constructed as a cosine transform of the function \(f\) as in the previous theorem. On the other hand, a new technique to construct positive definite functions from multiply monotone functions as well as from absolute monotone functions is presented and illustrated with several examples. Notice that a class of functions for which the interpolation problem in the multivariate case is uniquely solvable for any distinct point set \(A\subset \mathbb{R}^d,\) is the class of strictly positive definite functions.
0 references
multiply monotone functions
0 references
radialbasis functions
0 references
completely monotinic functions
0 references
interpolation of scattered data
0 references
0 references
0 references
0 references
0 references
0 references