We propose an extension to Markov Chain Monte Carlo methods for inferences in the imprecise probability framework. The algorithm is based on simultaneous sampling from all the Markov chains targeting the distributions in the credal set. The algorithm constructs a chain of random sets, which can be used for conservative estimation of lower and upper expected values of derived random variables. Tight bounds on the set of estimators arising from the set of admitted stochastic models can be obtained when the credal set is finite for general models. Conservative bounds can be obtained for some classes of models also when the credal set is uncountable. Computational complexity for uncountable credal sets is not bounded, and heuristic fixes need to be implemented.
Simultaneous Sampling for Robust Markov Chain Monte Carlo Inference
Space Technol.Proceedings
International Conference on Uncertainty Quantification & Optimisation ; 2020 ; Brussels, Belgium November 17, 2020 - November 20, 2020
Advances in Uncertainty Quantification and Optimization Under Uncertainty with Aerospace Applications ; Kapitel : 11 ; 173-183
2021-07-16
11 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Fitting multiple projective models using clustering-based Markov chain Monte Carlo inference
British Library Online Contents | 2015
|Markov Chain Monte Carlo Modular Ensemble Tracking
British Library Online Contents | 2013
|Sensitivity Analysis of Markov Chain Monte Carlo
British Library Conference Proceedings | 2010
|Sensitivity Analysis of Markov Chain Monte Carlo
AIAA | 2010
|