Optimal adaptive computations in the Jaffard algebra and localized frames (Q2654453): Difference between revisions

The authors study the numerical solution of infinite matrix equations \(\mathbf{A}\mathbf{u}=\mathbf{f}\) where the elements \(a_{jk}\) of \(\mathbf{A}\) satisfy \(|a_{jk}|\leq C(1+|j-k|)^{-\gamma}\), \(j,k\in \mathbb Z^d\) and \(\gamma > d\) (i.e., \(\mathbf {A}\) belongs to the Jaffard algebra). This kind of system of equations arises via frame discretizations using localized frames (e.g., \(|\langle g_j,g_k\rangle| \leq C'(1+|j-k|)^{-\gamma}\)). The main building blocks of the numerical algorithm is the approximation of infinite dimensional vectors and of bi-infinite matrix-vector products by finite ones and using these the authors obtain an adaptive numerical scheme that converges with quasi-optimal order and operation count. As an application the canonical dual frame of a localized frame is approximated and the approximations are shown to be frames. One key property used in the proofs is that the Jaffard algebra is closed under taking inverses.