Please use this identifier to cite or link to this item: https://scholarbank.nus.edu.sg/handle/10635/132131
DC FieldValue
dc.titlePRACTICAL INVESTIGATIONS ON BAYESIAN INVERSE PROBLEMS
dc.contributor.authorMUZAFFER EGE ALPER
dc.date.accessioned2016-11-30T18:00:44Z
dc.date.available2016-11-30T18:00:44Z
dc.date.issued2016-07-05
dc.identifier.citationMUZAFFER EGE ALPER (2016-07-05). PRACTICAL INVESTIGATIONS ON BAYESIAN INVERSE PROBLEMS. ScholarBank@NUS Repository.
dc.identifier.urihttp://scholarbank.nus.edu.sg/handle/10635/132131
dc.description.abstractInverse problems make up a challenging and practically important class of inference problems. Classical methods provide point estimates and confidence intervals which are asymptotically justified. As the computational power increased, however, ractitioners and researchers looked for better uncertainty quantification. The usual asymptotic confidence intervals gave way to full distributions using the Bayesian approach. This approach to inverse problems, while providing a full posterior distribution instead of a single point estimate as its answer, is also computationally much more expensive. Our contributions are two-fold. We present a novel adaptive sequential Monte Carlo method and its application to the groundwater-flow problem. Here, we observe significant time-savings compared to previous SMC approaches. We also observe, however, that this method is still too slow to be used in practice. Therefore, next, we turn our attention to multi-resolution (also known as multi-level) methods. We describe our implementation of this idea and show the match of experimental results to the predictions of asymptotic theory. This approach is promising for practical uncertainty quantification applications.
dc.language.isoen
dc.subjectInverse Problems, Sequential Monte Carlo, Bayesian Inference, Computational Statistics, Adaptive Monte Carlo, Multilevel Monte Carlo
dc.typeThesis
dc.contributor.departmentSTATISTICS & APPLIED PROBABILITY
dc.contributor.supervisorTHIERY, ALEXANDRE HOANG
dc.contributor.supervisorALEXANDROS BESKOS
dc.description.degreePh.D
dc.description.degreeconferredDOCTOR OF PHILOSOPHY
dc.identifier.isiutNOT_IN_WOS
Appears in Collections:Ph.D Theses (Open)

Show simple item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
thesis-final-signed-NOV.pdf6.35 MBAdobe PDF

OPEN

NoneView/Download

Page view(s)

149
checked on May 29, 2020

Download(s)

109
checked on May 29, 2020

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.