On the computation of spherical designs by a new optimization approach based on fast spherical Fourier transforms (Q652262): Difference between revisions

Let \({\mathbb S} \subset {\mathbb R}^3\) be the unit sphere. A \textit{spherical \(t\)-design} \(X_M = \{x_1,\ldots,x_M\} \subset \mathbb S\) provides a quadrature rule with equal weights for \(\mathbb S\), which is exact for polynomials up to degree \(t\). In this interesting paper, the authors consider the problem of numerical finding spherical \(t\)-designs for high degree \(t\). Therefore the authors compute local minimizers of a certain quadrature error \(A_t(X_M)\), which was introduced by \textit{I. H. Sloan} and \textit{R. S. Womersley} [J. Approx. Theory 159, No. 2, 308--318 (2009; Zbl 1183.05015)]. For the minimization problem, the authors regard several nonlinear optimization methods on Riemannian manifolds, like Newton and conjugate gradient methods. Using nonequispaced fast spherical Fourier transform (see \textit{S. Kunis} and \textit{D. Potts} [J. Comput. Appl. Math. 161, No. 1, 75--98 (2003; Zbl 1033.65123)]), the gradient and Hessian evaluations are performed in \({\mathcal O}(t^2\,\log t + M\,(\log \varepsilon)^2)\) arithmetic operations, where \(\varepsilon >0\) is a prescribed accuracy. Finally, special spherical \(t\)-designs for \(t\in \{60,\,100,\, 200,\,500,\,1000\}\) are presented, even in the case \(M\approx t^2/2\).