Quantification of uncertainty in high-dimensional stochastic problems using adaptive sparse grids

John Jakeman (The Australian National University)

MSI

DATE: 2010-07-05
TIME: 16:00:00 - 17:00:00
LOCATION: G35 John Dedman Building
CONTACT: JavaScript must be enabled to display this email address.

ABSTRACT:
Sparse grids have been frequently used for interpolation and integration of high-dimensional functions. However sparse grids can also be used to efficiently quantify uncertainty in high-dimensional stochastic problems. Sparse grids are extremely useful when only a small number of interactions between function variables contribute significantly to the system response. In such cases, which often occur in practice, high rates of convergence can be obtained even for highly-nonlinear models. In this talk I will present two forms of adaptive sparse grid algorithms that utilise local smoothness and low-effective dimensionality. Analytical and numerical convergence will be demonstrated and a number of numerical examples will be presented.
BIO:
http://wwwmaths.anu.edu.au/~jakeman