Department of Mathematics
Indian Institute of Science
Bangalore 560 012
GOLDEN JUBILEE YEAR 2005 - 2006
SEMINAR.....
Speaker |
: |
Prof. N. Hemachandra
|
Affiliation |
: |
Industrial Engg. and Operations Research, IIT, Mumbai
|
Subject Area |
: |
Mathematics
|
Venue |
: |
Lecture Hall I, Dept of Mathematics
|
Time |
: |
4.00 pm
|
Date |
: |
Thursday, 17th August 2006
|
Event Title |
: |
Sensitivity analysis and optimal ultimately stationary policies in some Markov decision models
|
Abstract | : | Consider a finite state-action constrained Markov decision model where both the objective and constraint functionals capture discounted costs. It is well known that optimal policies for such models usually are Markov but randomized. We first study sensitivity of such optimal policies w.r.t. to the upper bound vector of the constraint functionals. Feinberg-Shwartz (Math. of Operations Research, 1996) showed optimality of policies that are ultimately stationary deterministic (USD) for these models. They also proposed an iterative algorithm that involves linear programmes whose dimensions become progressively larger. We show how the above sensitivity analysis helps us in significantly improving their algorithm. |