Extrapolation and local acceleration of an iterative process for common fixed point problems (Q442508)
From MaRDI portal
| This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use this page instead for the normal view: Extrapolation and local acceleration of an iterative process for common fixed point problems |
scientific article; zbMATH DE number 6062866
| Language | Label | Description | Also known as |
|---|---|---|---|
| English | Extrapolation and local acceleration of an iterative process for common fixed point problems |
scientific article; zbMATH DE number 6062866 |
Statements
Extrapolation and local acceleration of an iterative process for common fixed point problems (English)
0 references
1 August 2012
0 references
common fixed points
0 references
cyclic projection method
0 references
cutter operator
0 references
quasi-nonexpansive operator
0 references
DoS Santos local acceleration
0 references
iterative process
0 references
Hilbert space
0 references
Goebel-Reich operators
0 references
maximal monotone operators
0 references
Crombez operators
0 references
local acceleration schemes
0 references
0 references
The article deals with iterative processes for common fixed point problems including so-called cutter operators. An operator \(T:\;H \to H\) in a Hilbert space \(H\) is called a cutter operator (sometimes firmly quasi-nonexpansive or directed or even separating one) if \(\text{Fix}\, T \subseteq H(x,Tx)\) for all \(x \in H\); here \(H(x,y) = \{u \in H|\;(u - y,x - y) \leq 0\}\). The class of cutter operators includes Goebel-Reich operators (\(\|Tx - Ty\|^2 \leq (Tx - Ty,x - y)\), \(x, y \in H\)), resolvents of maximal monotone operators, orthogonal and subgradient projections and others; in its turn this class is contained in the class of Crombez (or paracontracting) operators (\(\|Tx - q\| \leq \|x - q\|\), \(q \in \text{Fix}\, T\) and \(x \in H\)). The main part of the article deals with a finite family of cutter operators \(U_i:\;H \to H\), \(i = 1,\ldots,m\), with \(\bigcap_{i=1}^m \text{Fix}\, U_i \neq \emptyset\). The authors consider the iteration \(x^{k+1} = U_{\sigma,\lambda_k}x^k\) of the operator NEWLINE\[NEWLINEU_{\sigma,\lambda}x = x + \lambda\sigma(x)(Ux - x), \quad U = U_mU_{m-1} \cdots U_1,NEWLINE\]NEWLINE where \(\lambda \in (0,2)\) is a relaxation parameter, and \(\sigma:\;H \to (0,+\infty)\) is a step size function. The main theorem describes conditions under that the sequence of iterates \(x^k\) weakly converges to a fixed point of \(U\). The general result unifies and generalizes several existing local acceleration schemes. In particular, local acceleration of the sequential Kaczmarz method for linear equations, local acceleration of the sequential cyclic projection method for linear inequalities, and local acceleration of the sequential cyclic subgradient projection method are considered in the article.
0 references