ebook img

Iterative methods for ill-posed problems : an introduction PDF

150 Pages·2011·1.248 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Iterative methods for ill-posed problems : an introduction

Inverse and Ill-Posed Problems Series 54 Managing Editor Sergey I. Kabanikhin, Novosibirsk, Russia/Almaty, Kazakhstan Anatoly B. Bakushinsky Mihail Yu. Kokurin Alexandra Smirnova Iterative Methods for Ill-Posed Problems An Introduction De Gruyter MathematicsSubjectClassification2010:Primary:47A52;Secondary:65J20. ISBN 978-3-11-025064-0 e-ISBN 978-3-11-025065-7 ISSN 1381-4524 LibraryofCongressCataloging-in-PublicationData Bakushinskii,A.B.(AnatoliiBorisovich). [Iterativnyemetodyresheniianekorrektnykhzadach.English] Iterative methods for ill-posed problems : an introduction / by AnatolyBakushinsky,MikhailKokurin,AlexandraSmirnova. p.cm.(cid:2)(Inverseandill-posedproblemsseries;54) Includesbibliographicalreferencesandindex. ISBN978-3-11-025064-0(alk.paper) 1. Differential equations, Partial (cid:2) Improperly posed problems. 2.Iterativemethods(Mathematics) I.Kokurin,M.IU.(Mikhail IUr’evich) II.Smirnova,A.B.(AleksandraBorisovna) III.Title. QA377.B25513 2011 5151.353(cid:2)dc22 2010038154 BibliographicinformationpublishedbytheDeutscheNationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailedbibliographicdataareavailableintheInternetathttp://dnb.d-nb.de. ”2011WalterdeGruyterGmbH&Co.KG,Berlin/NewYork Typesetting:Da-TeXGerdBlumenstein,Leipzig,www.da-tex.de Printingandbinding:Hubert&Co.GmbH&Co.KG,Göttingen (cid:2)(cid:2)Printedonacid-freepaper PrintedinGermany www.degruyter.com Preface A variety of processes in science and engineering is commonly modeled by alge- braic, differential, integral and other equations. In a more difficult case, it can be systemsofequationscombinedwiththeassociatedinitialandboundaryconditions. Frequently,thestudyofappliedoptimizationproblemsisalsoreducedtosolvingthe corresponding equations. Typical examples include Euler’s equation in calculus of variationsandboundaryvalueproblemsforPontrjagin’smaximalprincipleincon- trol theory. All such equations, encountered both in theoretical and applied areas, may naturallybe classifiedasoperator equations. These equations connect the un- known parameters of the model with some given quantities describing the model. Theabovequantities,whichcanbeeithermeasuredorcalculatedatthepreliminary stage,formtheso-calledinputdata.Generally,theinputdataaswellastheunknown parametersaretheelementsofcertainmetricspaces,inparticular,BanachorHilbert spaces, with the operator of the model acting from the solution space to the data space.Thecurrenttextbookwillfocusoniterativemethodsforoperatorequationsin Hilbertspaces. Iterativemethodsintheirsimplestformarefirstintroducedinanundergraduate numericalanalysiscourse,amongwhichNewton’smethodforapproximatingaroot ofadifferentiablefunctioninonevariableisprobablythebestknown.Thisisatyp- icaliterativeprocesswidelyusedinapplications.Itcanbegeneralizedtothecaseof finitesystemsofnonlinearequationswithafinitenumberofunknowns,andalsoto thecaseofoperatorequationsininfinitedimensionalspaces.Itshould,however,be notedthatdirectgeneralizationofthiskindisonlypossibleforregularoperatorequa- tionsandsystemsofequations.Theregularityconditiongeneralizestherequirement onthederivativetobedifferentfromzeroinaneighborhoodoftheroot.Thisrequire- mentisusedfortheconvergenceanalysisofNewton’sschemeinaone-dimensional case.Withouttheregularitycondition,Newton’siterationsarenotnecessarilywell- defined.Thelackofregularityisamajorobstaclewhenitcomestoapplicabilityof notonlytheNewtonmethod,butallclassicaliterativemethods,gradient-typemeth- odsforexample,althoughoftenthesemethodsareformallyexecutableforirregular problems as well. Still, a lot of important mathematical models give rise to either irregular operator equations or to operator equations whose regularity is extremely vi Preface difficult to investigate, for instance numerous nonlinear inverse problems in PDEs. Thus,thequestioniswhetherornotitispossibletoconstructiterativemethodsfor nonlinearoperatorequationswithouttheregularitycondition. In the last few years the authors have been developing a unified approach to theconstructionofsuchmethodsforirregularequations.Theapproachunderdevel- opment is closely related to modern theory of ill-posed problems. The goal of our textbookistogiveabriefaccountofthisapproach.Thereare16chapters(lectures) in the manuscript, which is based on the lecture notes prepared by the authors for graduate students at Moscow Institute of Physics and Technology and Mari State University,Russia,andGeorgiaStateUniversity,USA.Asetofexercisesappearsat the end of each chapter. These range from routine tests of comprehension to more challenging problems helping to get a working understanding of the material. The bookdoesnotrequireanypriorknowledgeofclassicaliterativemethodsfornonlin- ear operator equations. The first three chapters investigate the basic iterative meth- ods,theNewton,theGauss–Newtonandthegradientones,ingreatdetail.Theyalso give an overview of some relevant functional analysis and infinite dimensional op- timization theory. Further chapters gradually take the reader to the area of iterative methods for irregular operator equations. The last three chapters contain a number ofrealisticnonlineartestproblemsreducedtofinitesystemsofnonlinearequations withafinitenumberofunknowns,integralequationsofthefirstkind,andparameter identificationproblemsinPDEs.Thetestproblemsarespeciallyselectedinorderto emphasize numerical implementation of various iteratively regularized procedures addressed in this book, and to enable the reader to conduct his/her own computa- tionalexperiments. Asitfollowsfromthetitle,thistextbookismeanttoilluminateonlytheprimary approachestotheconstructionandinvestigationofiterativemethodsforsolvingill- posed operator equations. These methods are being constantly perfected and aug- mentedwithnewalgorithms.Appliedinverseproblemsarethemainsourcesofthis development:tosolvethem,thesuccessfulimplementationofwell-knowntheoreti- calproceduresisoftenimpossiblewithoutadeepanalysisofthenatureofaproblem andasuccessfulresolutionofthedifficultiesrelatedtothechoiceofcontrolparame- ters,whichsometimesnecessitatesmodificationoftheoriginaliterativeschemes.At times,byanalyzingthestructureofparticularappliedproblems,researchersdevelop new procedures (iterative algorithms, for instance), aimed at these problems exclu- sively.Thenew‘problem-oriented’proceduresmayturnouttobemoreeffectivethan thosedesignedforgeneraloperatorequations.Examplesofsuchproceduresinclude, but are not limited to, the method of quasi-reversibility (Lattes and Lions, 1967) for solving unstable initial value problems (IVPs) for the diffusion equation with reversed time, iteratively regularized schemes for solving unstable boundary value problems (BVPs), which reduce the original BVP to a sequence of auxiliary BVPs for the same differential equation with ‘regularized’ boundary conditions (Kozlov and Mazya, 1990), and various procedures for solving inverse scattering problems. For applied problems of shape design and shape recovery, the level set method is widelyused(OsherandSethian,1988).Thereadermayconsult[59,27,63,69,40] foradetailedtheoreticalandnumericalanalysisofthesealgorithms. Preface vii The formulas within the text are doubly numbered, with the first number being thenumberofthechapterandthesecondnumberbeingthenumberoftheformula within the chapter. The problems are doubly numbered as well. A few references aregiventotheextensivebibliographyattheendofthebook;theyareindicatedby initials in square brackets. Standard notations are used throughout the book; R is the set of real numbers, N is the set of natural numbers. All other notations are introducedastheyappear. Theauthorshopethatthetextbookwillbeusefultograduatestudentspursuing their degrees in computational and applied mathematics, as well as to researchers andengineerswhomayencounternumericalmethodsfornonlinearmodelsintheir work. AnatolyBakushinsky MikhailKokurin AlexandraSmirnova Contents Preface............................................................ v 1 Theregularitycondition.Newton’smethod ....................... 1 1.1 Preliminaryresults ......................................... 1 1.2 Linearizationprocedure ..................................... 2 1.3 Erroranalysis.............................................. 4 Problems ...................................................... 6 2 TheGauss–Newtonmethod ..................................... 10 2.1 Motivation ................................................ 10 2.2 Convergencerates .......................................... 12 Problems ...................................................... 14 3 Thegradientmethod ........................................... 16 3.1 Thegradientmethodforregularproblems ...................... 16 3.2 Ill-posedcase.............................................. 18 Problems ...................................................... 20 4 Tikhonov’sscheme ............................................. 23 4.1 TheTikhonovfunctional .................................... 23 4.2 Propertiesofaminimizingsequence........................... 24 4.3 Othertypesofconvergence .................................. 27 4.4 Equationswithnoisydata.................................... 29 Problems ...................................................... 30 5 Tikhonov’sschemeforlinearequations........................... 32 5.1 Themainconvergenceresult ................................. 32 5.2 Elementsofspectraltheory .................................. 34 5.3 Minimizingsequencesforlinearequations ..................... 35 5.4 Aprioriagreementbetweentheregularizationparameterandthe errorforequationswithperturbedright-handsides............... 37

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.