Quasi-Monte Carlo sampling in multilevel, multiindex and multivariate-decomposition methods

Dirk Nuyens (KU Leuven)

May 05. 2020, 16:35 — 17:15

The aim of this talk is to explain what the multivariate-decomposition method is without going into too many technical details. I will introduce the topic of infinite-dimensional integration starting from a model problem for uncertainty quantification. Then I will spend some time on the multilevel method and the typical error bound statements that can be found in the literature. I will take the liberty to also present an atypical statement for deterministic quasi-Monte Carlo rules to set the scene for the multivariate-decomposition method which I will discuss next. As a final method I will touch the multiindex method and then finally end with some numerical results.
 

 

Further Information
Venue:
Erwin Schrödinger Institute - virtual
Associated Event:
Multilevel and multifidelity sampling methods in UQ for PDEs (Online Workshop)
Organizer(s):
Kody Law (U Manchester)
Fabio Nobile (EPFL Lausanne)
Robert Scheichl (U Heidelberg)
Karen Willcox (U of Texas, Austin)