ACMD Seminar: Information-theoretic tools for uncertainty quantification of high dimensional stochastic models

Petr Plechac
Department of Mathematical Sciences, University of Delaware

Friday, July 28, 2017, 15:00 - 16:00
Building 101, Lecture Room F
Gaithersburg

Friday, July 28, 2017, 13:00 - 14:00
Room 1-4058
Boulder

Host: Michael Mascagni

Abstract: We present mathematical tools for deriving optimal, computable bounds on sensitivity indices of observables for complex stochastic models arising in biology, reaction kinetics and materials science. The presented technique allows for deriving bounds also for path-dependent functionals and risk sensitive functionals. Using the rate of relative entropy the sensitivity of a wide class of observables can be bounded by Fisher information and quantities that characterize the statistics (variance, autocorrelation) of observables. The use of variational representation of relative entropy also allows for error estimation and uncertainty quantification in coarse-grained models.

Bio: Petr Plechac is Professor of Applied Mathematics in the Department of Mathematical Sciences, the University of Delaware. Before coming to UD he was associate professor at Warwick University, United Kingdom and then held joint faculty appointment as associate professor and senior research scientist at the University of Tennessee, Knoxville and the Oak Ridge National Laboratory. Plechac earned PhD in numerical analysis from Charles University, Prague, the Czech Republic and gained postdoctoral experience at Oxford University.

His research focuses primarily on computational methods and numerical analysis of algorithms for problems arising in materials modeling, chemical reaction kinetics, electronic structure computations and stochastic simulations of molecular systems.  He has developed computational techniques for  multi-scale simulations of materials microstructure, reaction kinetics and Monte Carlo sampling in molecular systems. In his recent work he has contributed to mathematical development of new information-theoretic tools for error estimation, coarse-graining, model reduction and uncertainty quantification in high-dimensional stochastic systems. His research has been funded by the National Science Fundation, the U.S. Department of Energy and the U.S. Department of Defense.

Plechac serves as an associate editor of SIAM Journal on Numerical Analysis and SIAM  Multiscale Modeling and Simulations.

Note: Visitors from outside NIST must contact Cathy Graham; (301) 975-3800; at least 24 hours in advance.

Part of the ACMD Seminar Series.