Uncertainty Quantification Working Group
March 21, 11:30 AM, CNLS Conf. Room, TA-3, Bldg. 1690

Novel Uncertainty Quantification Methods Based on Generalized Information Theory

Cliff Joslyn, Modeling, Algorithms, and Informatics, CCS-3

Abstract

Researchers at the Los Alamos and Sandia National Laboratories are exploring the utility of non-traditional methods of uncertainty quantitifcaion for engineering modeling applications such as risk and reliability analysis. Described broadly as Generalized Information Theory (GIT), these approaches include random set theory, Dempster-Shafer evidence theory, interval analysis and probability intervals, possibility theory, and fuzzy systems, amongst others. Such approaches are intended to complement and augment statistical, Bayesian, and traditional probabilistic methods of uncertainty quantification when input data are highly sparse, qualitative, imprecise, or derived from non-quantitative sources such as linguistic expressions.

A workshop held at Los Alamos at the end of February (http://www.c3.lanl.gov/~joslyn/epistemic) provided a forum for the LANL research community and members of Sandia's Epistemic Uncertainty Project to share and discuss their technical approaches in this area. In this talk we will first present a synoptic discussion of the mathematical formalism of the components of GIT mentioned above, and more importantly their formal relations. We will then briefly review a number of the presentations from the workhsop and the work being done for the ASCI program in the quantification of epistemic uncertainty in large numerical simulations (http://www.sandia.gov/epistemic).

To send e-mail to author: joslyn@lanl.gov . His web site is http://www.c3.lanl.gov/~joslyn/.