Font Size: a A A

Regularization of quasi-Newton methods: Applied to image restoration

Posted on:2005-11-23Degree:Ph.DType:Thesis
University:Emory UniversityCandidate:Palmer, Katrina MarieFull Text:PDF
GTID:2450390008993966Subject:Mathematics
Abstract/Summary:
Quasi-Newton methods have been widely studied for well-posed problems. This thesis provides more insight to the regularization properties of quasi-Newton methods; that is, the behavior of quasi-Newton methods on ill-posed problems. The four quasi-Newton methods considered are conjugate gradient least squares (CGLS), Barzilai-Borwein (BB), residual norm steepest descent (RNSD) and Landweber (LW). These regularization properties are studied in two ways. The first is by analyzing the so-called "filter factors" of the four names quasi-Newton methods. The second was is by observing the behavior of the residual on CGLS and RNSD.
Keywords/Search Tags:Quasi-newton methods, Regularization
Related items