Bibliographic Metadata

Title
A semismooth Newton method for L1 data fitting with automatic choice of regularization parameters and noise calibration
AuthorKunisch, Karl In der Gemeinsamen Normdatei der DNB nachschlagen ; Clason, Christian ; Jin, Bangti
Published in
SIAM Journal on Imaging Sciences, Philadelphia, Pa., 2010, Vol. 3, Issue 2, page 199-231
PublishedSIAM
Edition
Publisher version
LanguageEnglish
Document typeJournal Article
Keywords (EN)L1 data fitting / semismooth Newton / Fenchel duality / regularization parameter / balancing principle / model function
URNurn:nbn:at:at-ubg:3-189 Persistent Identifier (URN)
DOIdoi:10.1137/090758003 
Restriction-Information
 The work is publicly available
Files
A semismooth Newton method for L1 data fitting with automatic choice of regularization parameters and noise calibration [1.87 mb]
Links
Reference
Classification
Abstract (English)

This paper considers the numerical solution of inverse problems with an L1 data fitting term, whichis challenging due to the lack of differentiability of the objective functional. Utilizing convex duality,the problem is reformulated as minimizing a smooth functional with pointwise constraints,which can be efficiently solved using a semismooth Newton method. In order to achieve superlinearconvergence, the dual problem requires additional regularization. For both the primal and the dualproblems, the choice of the regularization parameters is crucial. We propose adaptive strategiesfor choosing these parameters. The regularization parameter in the primal formulation is chosenaccording to a balancing principle derived from the model function approach, whereas the one in thedual formulation is determined by a path-following strategy based on the structure of the optimalityconditions. Several numerical experiments confirm the efficiency and robustness of the proposedmethod and adaptive strategy.

Notice