Fixed Point Algorithms for Nonconvex Feasibility with Applications

by Robert Hesse
Doctoral thesis
Date of Examination:2014-07-14
Date of issue:2014-07-31
Advisor:Prof. Dr. Russell Luke
Referee:Prof. Dr. Russell Luke
Referee:Prof. Dr. Gerlind Plonka-Hoch
Persistent Address: http://dx.doi.org/10.53846/goediss-4623

 

 

Files in this item

Name:SUBlinearized.pdf
Size:1.44Mb
Format:PDF

The following license files are associated with this item:

Abstract

English

Projection algorithms for solving (nonconvex) feasibility problems provide powerful and computationally efficient schemes for a wide variety of applications. Algorithms as Alternating Projections (AP) and Douglas-Rachford (DR) are two of the more prominent projection algorithms in imaging sciences and signal processing. A nonconvex framework is introduced that enables a general and new approach to characterizing the convergence behavior of general fixed point operators.  In classical fixed point theory, firm nonexpansiveness of mappings is a property that is often used to show convergence of a broad class of algorithms. As our main interest is dealing with nonconvex feasibility, the described methods no longer match the notion of firm nonexpansiveness. The framework, theorems and concepts developed in this work generalize the tools from convex analysis for the analysis of fixed-point iterations of operators that violate the classical property of firm nonexpansiveness in some quantifiable fashion.  Using these techniques a convergence analysis on AP and DR is carried out. The analysis is then applied to the problem of phase retrieval and ptychographic imaging.
Keywords: Algorithms; Douglas-Rachford; Alternating Projections; Phase Retrieval; Ptychography; metric regularity; linear regularity; uniform regularity; set regularity; regularity of collections of sets; linear convergence; feasibility problems; firmly nonexpansive mappings
 
Statistik