The equivalence between \(T\)-stabilities of the Krasnoselskij and the Mann iterations (Q2472303): Difference between revisions

The equivalence mentioned in the title is proved for the following iterations: \[ \begin{aligned} e_{n+1}&=(1-\alpha_n)e_n+\alpha_nTe_n\;\text{ (Mann),}\\ p_{n+1}&= (1-\lambda)p_n+\lambda Tp_n\;\text{ (Krasnoselskii)}, \end{aligned} \] where \(T:X\to X\) is a selfmap of a normed space. \(T\)-stability means that \(\varepsilon_n\to 0\) (resp., \(\delta_n\to 0\)) implies convergence of the iteration, where \(\varepsilon_n=\| e_{n+1}-(1-\alpha_n)e_n-\alpha_nTe_n\| \) (resp., \(\delta_n=\| p_{n+1}-(1-\lambda)p_n-\lambda Tp_n\|\)). The equivalence is checked in the case \(\alpha_n\to \lambda\in (0,1)\) and \(T\) with bounded range, or for sequences with \(| | v_n-Tv_n| | \to 0\). Some comments on how these methods compare with the Picard--Banach iterations are given in the last section of the paper.