scientific article; zbMATH DE number 1329446

From MaRDI portal
Publication:4257764





The author presents a general regularization scheme for computing a solution \(x^* \in D(F)\subset X\) of a nonlinear ill-posed problem \(F(x)=y\), where \(F: D(F)\subset X \rightarrow Y\) is a nonlinear operator between \(X\) and \(Y\) and instead of \(y\) noisy data \(y^{\delta}\in Y\) with \(\|y-y^{\delta}\|\leq \delta\) are given. In the suggested scheme regularized solutions \(x^{\delta}_{\alpha}\) are defined as a fixed point of the following nonlinear operator \[ Tx=\overline{x}+g_{\alpha}(B^{-s}F^{'*}F')B^{-s}F^{'*}\{y^{\delta}-F(x)-F'( \overline{x}-x)\}, \] where \(\overline{x}\in D(F)\) is an initial guess, \(F'=F'(x)\), and \(B\) denotes an unbounded selfadjoint strictly positive operator. Suppose \(\|x\|_r=\|B^{r/2}x\|\) is the norm of \(x\in X\) in a Hilbert scale \((X_r)_{r\in \mathbb{R}}\). Assuming \(\|F'(x_0)x\|\sim \|x\|_{-a}\) for \(x_0 \in B_r(x^*)\) and \(\|x^*-\overline{x}\|_p \leq E\) for some \(a>0\) and \(p\geq 0\) the author proves that the fixed point \(x^{\delta}_{\alpha}\) satisfies the order optimal error bound \(\|x^{\delta}_{\alpha}-x^*\|\leq c \delta^{p/(p+\alpha)}\) provided that \(g_{\alpha}\) and the regularization parameter \(\alpha\) are chosen properly.











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 Q4257764)