The Bayesian Approach to Inverse Problems
 Creators
 Dashti, Masoumeh
 Stuart, Andrew M.
 Others:
 Ghanem, Roger
 Higdon, David
 Owhadi, Houman
Abstract
These lecture notes highlight the mathematical and computational structure relating to the formulation of, and development of algorithms for, the Bayesian approach to inverse problems in differential equations. This approach is fundamental in the quantification of uncertainty within applications in volving the blending of mathematical models with data. The finite dimensional situation is described first, along with some motivational examples. Then the development of probability measures on separable Banach space is undertaken, using a random series over an infinite set of functions to construct draws; these probability measures are used as priors in the Bayesian approach to inverse problems. Regularity of draws from the priors is studied in the natural Sobolev or Besov spaces implied by the choice of functions in the random series construction, and the Kolmogorov continuity theorem is used to extend regularity considerations to the space of Hölder continuous functions. Bayes' theorem is de rived in this prior setting, and here interpreted as finding conditions under which the posterior is absolutely continuous with respect to the prior, and determining a formula for the RadonNikodym derivative in terms of the likelihood of the data. Having established the form of the posterior, we then describe various properties common to it in the infinite dimensional setting. These properties include wellposedness, approximation theory, and the existence of maximum a posteriori estimators. We then describe measurepreserving dynamics, again on the infinite dimensional space, including Markov chainMonte C arlo and sequential Monte Carlo methods, and measurepreserving reversible stochastic differential equations. By formulating the theory and algorithms on the underlying infinite dimensional space, we obtain a framework suitable for rigorous analysis of the accuracy of reconstructions, of computational complexity, as well as naturally constructing algorithms which perform well under mesh refinement, since they are inherently welldefined in infinite dimensions.
Additional Information
© 2017 Springer. The authors are indebted to Martin Hairer for help in the development of these notes, and in particular for considerable help in structuring the Appendix, for the proof of Theorem 28 (which is a slight generalization to Hilbert scales of Theorem 6.16 in [40]) and for the proof of Corollary 5 (which is a generalization of Corollary 3.22 in [40] to the nonGaussian setting and to Hölder, rather than Lipschitz, functions {Ψ_k}). They are also grateful to Joris Bierkens, Patrick Conrad, Matthew Dunlop, Shiwei Lan, Yulong Lu, Daniel SanzAlonso, Claudia Schillings and Aretha Teckentrup for careful proofreading of the notes and related comments. AMS is grateful for various hosts who gave him the opportunity to teach this material in short course form at TIFRBangalore (Amit Apte), Göttingen (Axel Munk), PKUBeijing (Teijun Li), ETHZurich (Christoph Schwab) and Cambridge CCA (Arieh Iserles), a process which led to refinements of the material; the authors are also grateful to the students on those courses, who provided useful feedback. The authors would also like to thank Sergios Agapiou and YuanXiang Zhang for help in the preparation of these lecture notes, including typesetting, proofreading, providing the proof of Lemma 3and delivering problems classes related to the short courses. AMS is also pleased to acknowledge the financial support of EPSRC, ERC and ONR over the last decade, while the research that underpins this work has been developed.
Attached Files
Submitted  1302.6989v4.pdf
Files
Name  Size  Download all 

md5:c31336c42f0f1be4fa2a77aba4b4925e

1.0 MB  Preview Download 
Additional details
 Eprint ID
 71941
 DOI
 10.1007/9783319123851_7
 Resolver ID
 CaltechAUTHORS:20161111104641272
 arXiv
 arXiv:1302.6989
 Engineering and Physical Sciences Research Council (EPSRC)
 European Research Council (ERC)
 Office of Naval Research (ONR)
 Created

20161115Created from EPrint's datestamp field
 Updated

20230606Created from EPrint's last_modified field