[Colloquium] Rocklin/MS Presentation/May 18, 2011

[Margaret Jaffey]

This is an announcement of Matthew Rocklin's MS Presentation.

------------------------------------------------------------------------------
Date:  Wednesday, May 18, 2011

Time:  12:00 PM

Place:  Ryerson 251

M.S. Candidate:  Matthew Rocklin

M.S. Paper Title: Uncertainty Quantification and Sensitivity Analysis

Abstract:
Dynamical systems describe the time evolution of a point in a space of
potential states. If knowledge about the initial point is uncertain
then the dynamical system evolves a probability distribution forward
rather than a single point. Tracking this distribution is challenging
and approximate methods are often necessary.

We study uncertainty quantification of short-term forecasts in
dynamical systems. We compare sampling (Monte Carlo) against adjoint
methods on low and high dimensional systems. For sampling we
particularly consider the formation of sensible initial distributions,
so-called ”errors of the day”, in a low-rank attractor given model
dynamics and expectation values distributed in time. Adjoint methods
are shown to be a cheaper alternative when close to the true solution
which can also be used to pinpoint optimal observations to reduce
targeted uncertainties in the final distribution.

For a motivating example we consider uncertainty in windspeed of
short-term forecasts to aid with advance management of resources on
the power grid.

Matthew's advisor is Prof. Todd Dupont

Login to the Computer Science Department website for details:
 https://www.cs.uchicago.edu/phd/ms_announcements#mrocklin

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Margaret P. Jaffey            margaret at cs.uchicago.edu
Department of Computer Science
Student Support Rep (Ry 156)               (773) 702-6011
The University of Chicago      http://www.cs.uchicago.edu
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=