Best approximation, Tikhonov regularization and reproducing kernels (Q2574400): Difference between revisions

This paper deals with approximation theory in reproducing kernel Hilbert spaces (RKHS) \(H_K\) admitting a reproducing kernel \(K(p,q)\) on some set \(E\). In particular, it yields some contributions to the analysis of least-squares solutions for linear operator equations \(Lf=d\), where \(L:H_K \to {\mathcal H}\) is a bounded linear operator mapping from the RKHS \(H_K\) to some Hilbert space \({\mathcal H}\). Problems of interest are such that the Moore-Penrose inverse of \(L\) is unbounded, i.e., linear ill-posed problems. Then instead of the conventional least-squares solution penalized least-squares problems are considered. This approach is known as Tikhonov regularization in inverse problems theory. Some error and stability results are given for the Tikhonov regularization in RKHS. Specific results are formulated if the operator \(LL^*\) is unitary and moreover if \(L\) is compact.