| |
Stochastic Dynamical Systems and Control |
| March 26, 2007 to March 30, 2007 |
| Organized By: Jonathan Mattingly (Duke), Igor Mezic (UCSB-Chair), Andrew Stuart (Warwick) |
| |
| Parent Programs: |
| Dynamical Systems |
| |
| Return to Workshop Description |
| |
A sample path optimality result for a class of Markov decision processes
Tuesday March 27, 2007
02:00PM - 03:00PM
Speakers:
Fern Hunt
Abstract:
In this talk we discuss a class of controlled stochastic processes
whose state process under any stationary policy is a uniformly ergodic
Markov chain. Here the state space is countable and the action space (the set of
controls)is finite. A policy that minimizes the long term expected average
cost can be computed using linear programming methods. For practical
applications it would be useful to know how well the optimal policy performs
on actual sample paths at least in some asymptotic sense. We present
sufficient conditions for a sample path optimality in this setting and if
time permits we will present the application problem that motivated this work.
|
