scientific article; zbMATH DE number 3888813

From MaRDI portal
Publication:3221289





The normally solvable equation (*) \(Au=f\) with a linear bounded operator \(A: H\to F\), \(H\), \(F\) being Hilbert spaces, is investigated. One supposes \(\| A-\tilde A\| <h\), \(\| f-\tilde f\| <\delta\) for \(h>0\), \(\delta >0\) and takes the regularization functional depending on \(\alpha =\alpha (h)>0\) such that \(\tilde u_{\alpha}=(\alpha E+\tilde A^*\tilde A)^{-1}\tilde A^*\tilde f\). Theorem 1: Let (*) be consistent and normally solvable, and denote by \(\bar u\) the exact solution of (*). Then there exists a dependence on \(\alpha =\alpha (h)>0\) such that \(\| \tilde u_{\alpha (h)}-\bar u\|_H=O(h+\delta)\). In the nonconsistent case the stable orthogonal projection \(Pf\) of the right-hand side \(f\) on \(R(A)\) using \(\tilde f_{\alpha}=(\alpha E+\tilde A\tilde A^*)^{-1}\tilde A\tilde A^*\tilde f\) is employed, which seems to be the regularization functional of Tikhonov for the consistent equation \(\tilde A^*g=\tilde A^*\tilde f\). Setting \(\alpha = h\), \(f_h\) approximates the right-hand side of the consistent equation \(Au=Pf\), and for the pseudo-solution \(\tilde u_h\) the same estimate as in Theorem 1 holds. [For the entire collection see Zbl 0536.00008.]












This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3221289)