Kernel based quadrature on spheres and other homogeneous spaces (Q2450862): Difference between revisions

The authors study the quadrature (or cubature) formulas which are associated with certain classes of positive definite and conditionally positive definite kernels that are invariant under the group action of a homogeneous manifold. New error estimates for interpolation/reproduction problems on two-point homogeneous manifolds are obtained. The quadrature formulas developed in this paper are for the general setting of a compact, homogeneous, \(n\)-dimensional manifold \(M\) that is equipped with a group invariant Riemannian metric \(g_{ij}\) and its associated invariant measure \(d\mu(x)=\sqrt{\det(g_{ij}(x))}dx\); namely, \[ \int_{M}f(x)d\mu(x)=\sum_{\xi\in X}c_{\xi}f(\xi)=:Q(f), f\in C(M). \] The finite set \(X\subset M\) is composed of centers/nodes, and the weights \(\{c_{\xi}\}_{\xi\in X}\) are chosen so that \(Q\) integrates exactly a given finite-dimensional space of continuous functions, \(V\). These quadrature formulas are stable both under an increase in the number of points and in the presence of noise. If the number of points is increased, then the norm of the quadrature operator remains uniformly bounded, as long as the level of quasi-uniformity is maintained. These quadrature formulas can be used to obtain similar formulas for manifolds diffeomorphic to \(S^n\), oblate spheroids for instance. The weights are obtained by solving a single linear system. For \(S^2\), and the restricted thin plate spline kernel \(r^2\log r\), these weights can be computed for two-thirds of a million nodes, using a preconditioned iterative technique.