Continuous and discrete least-squares approximation by radial basis functions on spheres (Q860692)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Continuous and discrete least-squares approximation by radial basis functions on spheres
scientific article

    Statements

    Continuous and discrete least-squares approximation by radial basis functions on spheres (English)
    0 references
    0 references
    0 references
    0 references
    0 references
    9 January 2007
    0 references
    Let \(S^n\) be the \(n\)-dimensional sphere in \(\mathbb R^n\) and \(A=\{U_j,\psi_j\}_{j=1}^m\) be an atlas for \(S^n\), i.e. open sets \(U_j\subset S^n\) cover \(S^n\), \(\psi_j\) are homeomorphic from \(U_j\) to the unit ball \(B(0,1)\subset \mathbb R^n\) and \(\psi_i\circ \psi_j^{-1}\) are \(C^{\infty}\) on \(\psi_j(U_j\cap U_i)\). Let \(\{\chi_j\}_{j=1}^m\) be the partition of unity with respect to \(\{U_j\}_{j=1}^m\). For \(f:S^n\to \mathbb R\) we introduce \(\pi_j(f)(x)=f\circ \psi_j^{-1}(x)X_{B(0,1)}(x)\), where \(X_E\) is the indicator of \(E\). Then \[ W^{\tau}_p(S^n)=\Biggr\{f\in L_p(S^n): \| f| W^{\tau}_p(S^n)\| = \biggl(\sum^m_{j=1}\| \pi_j(\chi_jf)| W^{\tau}_p(\mathbb R^n)\| ^p\biggr)^{1/p}<\infty\Biggr\}. \] There \(1\leq p<\infty\), for \(\tau\) see below. The authors consider concrete \(\psi_j\) similar to stereographic projections. Further \(h_{X, S^n}=\sup_{x\in S^n}min_{x_j\in X}d(x,x_j)\), where \(d(x,y)\) is the spherical distance and \(X\) is finite. Also \(\tau=k+s\), where \(k\in\mathbb N\), \(0\leq s<1\), \(1\leq p<\infty\), \(1\leq q\leq\infty\), \(m\in\mathbb N_0\) and \(k>m+n/p\) if \(p>1\) or \(k\geq m+n/p\) if \(p=1\). Theorem 3.3. Suppose that the assumption above holds, \(X\subset S^n\) is finite and \(h_{X,S^n}\) is sufficiently small. If \(u\in W^{\tau}_p(S^n)\) satisfies \(u| X=0\), then \[ \| u| W^{m}_q(S^n)\| \leq Ch^{\tau-m-n(1/p-1/q)_+}\| u| W^{\tau}_p(S^n)\| . \] Applications to estimates of approximation of \(u\) by solutions of some least-square problems are given.
    0 references
    0 references
    radial basis functions
    0 references
    error estimates
    0 references
    scattered data
    0 references
    norming sets
    0 references
    0 references