|Date:||May 7, 2018|
|Place:||Lecture Hall, Research I|
Abstract: Multifidelity approaches attempt to construct, for example, a bi-fidelity model having an accuracy comparable with the high-fidelity model and computational cost comparable with the low-fidelity model. Allocation decisions that distribute computational resources across several simulation models become extremely important in situations where only a small number of expensive high fidelity simulations can be run. In this talk, we present a novel approach based upon machine learning concepts to allocating resources for expensive simulations of high fidelity models when used in a multifidelity framework. We then present a bi-fidelity framework defined by temporal and/or spatial discretization parameters that relies on the low-rank structure of the map between model parameters/uncertain inputs and the solution of interest. We show how this framework behaves on canonical examples, and we show real-world (practical) extensions of our methodology to various scenarios from continuum-level modeling to molecular dynamics simulations.
The colloquium is preceded by tea from 16:45 in the Resnikoff Mathematics Common Room, Research I, 127.