Fixed point Ishikawa iterations (Q1206943): Difference between revisions

If \(J\) is the closed unit inverval, with \(T\) a selfmap of \(J\), the Ishikawa iterates of \(T\) are defined by \(u_{n+1}=[(1-\alpha_ n)u_ n+\alpha_ nT[(1-\beta_ n)u_ n+b_ n Tu_ n]]\) with \(u_ 0\in J\) and \(\{\alpha_ n\}\), \(\{\beta_ n\}\) satisfying the three conditions \[ \text{(a)} \quad 0\leq \alpha_ n\leq \beta_ n\leq 1, \qquad \text{(b)} \quad \lim \beta_ n=0, \qquad \text{(c)} \quad \sum \alpha_ n\beta_ n=\infty. \] In the special case \(\beta_ n=0\) \(\forall n\), \textit{W. R. Mann} [Proc. Am. Math. Soc. 4, 506-510 (1953; Zbl 0050.116)] showed that if \(T\) has a unique fixed point in \(J\), the iteration scheme converges to that fixed point. The second author [J. Math. Anal. Appl. 56, 741-750 (1976; Zbl 0353.47029)] extended the result, without the uniqueness requirement, to an Ishikawa process with the restrictions \[ \text{(i)} \quad \alpha_ n\geq 0,\;\beta_ n\leq 1, \qquad \text{(ii)} \quad \lim \alpha_ n=\lim \beta_ n=0, \qquad \text{(iii)} \quad \sum \alpha_ n=\infty. \] The authors of this paper show that the condition \(\lim \alpha_ n=0\) in (ii) can be removed. This result may be extended to Banach spaces.