New Douglas-Rachford algorithmic structures and their convergence analyses

Yair Censor,Rafiq Mansour
DOI: https://doi.org/10.48550/arXiv.1512.00409
2015-12-02
Abstract:In this paper we study new algorithmic structures with Douglas- Rachford (DR) operators to solve convex feasibility problems. We propose to embed the basic two-set-DR algorithmic operator into the String-Averaging Projections (SAP) and into the Block-Iterative Pro- jection (BIP) algorithmic structures, thereby creating new DR algo- rithmic schemes that include the recently proposed cyclic Douglas- Rachford algorithm and the averaged DR algorithm as special cases. We further propose and investigate a new multiple-set-DR algorithmic operator. Convergence of all these algorithmic schemes is studied by using properties of strongly quasi-nonexpansive operators and firmly nonexpansive operators.
Optimization and Control
What problem does this paper attempt to address?