Weak Convergence of Numerical Methods for Dynamical Systems and Optimal Control, and a relation with Large Deviations for Stochastic Equations
Thursday, October 21, 2010 - 8:30am - 9:30am
Mattias Sandberg (Royal Institute of Technology (KTH))
I will present a method to prove weak convergence of numerical methods for dynamical systems, using dual solutions. This general method is applied to optimal control problems, and is used to prove convergence of approximate value functions. The theory of large deviations will also be mentioned. It makes it possible to represent rare event solutions to stochastic differential equations as solutions of optimal control problems. This representation will be used on a particular stochastic partial differential equation arising in the study of phase transitions. It will be shown how the resulting optimal control problem can be analyzed, again with the same kind of method to prove weak convergence.