Reich's problem concerning Halpern's convergence (Q2390919): Difference between revisions

Let \(E\) be a linear normed space, \(C\) be a convex subset of \(E\) and let \(T:C\rightarrow C\) be a self-mapping of \(C\). In order to approximate fixed points of \(T\), several iterative schemes have been considered, e.g., Picard iteration, Krasnoselskij iteration, Mann iteration, Ishikawa iteration, etc., and their variants, see, for example, the reviewer's recent monograph [\textit{V.\,Berinde}, ``Iterative approximation of fixed points'' (First edition: Efemeride, Baia Mare) (Second Edition: Lecture Notes in Mathematics 1912, Springer, Berlin) (\({}^1\)2002; Zbl 1036.47037) (\({}^2\)2007; Zbl 1165.47047)]. The Mann iterative scheme \(\{x_n\}\) is defined by \[ x_{n+1}=(1-\alpha_n)x_n+\alpha_n\,T x_n,\quad n\geq 1, \] where \(x_1\in C\) and \(0\leq \alpha_n\leq 1,\) while Halpern's iterative scheme \(\{x_n\}\), although defined by a formula which is very similar to that defining Mann iteration, is slightly different to the latter, due to the fact that the first term of the convex combination is a fixed \(u\in C\), that is, \[ x_{n+1}=(1-\alpha_n)u+\alpha_n\,T x_n,\quad n\geq 1. \] In order to ensure convergence of the Halpern's iteration, several conditions on the parameter sequence \(\{\alpha_n\}\) have been considered in literature. The author of the paper under review collects nine such conditions, numbered \((C_1)\)--\((C_9)\), and proves, among others, that Wittmann's type condition \((C_9)\) is the weakest sufficient condition, while the conjunction of \((C_1)\) and \((C_2)\) is a sufficient and necessary condition for Halpern's iteration to be convergent. Several illustrative examples are given.