DocumentCode :
3311648
Title :
A simulation-based method for aggregating Markov chains
Author :
Deng, Kun ; Mehta, Prashant G. ; Meyn, Sean P.
Author_Institution :
Coordinated Sci. Lab., Univ. of Illinois at Urbana-Champaign, Urbana, IL, USA
fYear :
2009
fDate :
15-18 Dec. 2009
Firstpage :
4710
Lastpage :
4716
Abstract :
This paper addresses model reduction for a Markov chain on a large state space. A simulation-based framework is introduced to perform state aggregation of the Markov chain based on observations of a single sample path. The Kullback-Leibler (K-L) divergence rate is employed as a metric to measure the distance between two stationary Markov chains. Model reduction with respect to this metric is cast as an infinite-horizon average cost optimal control problem. In this way an optimal policy corresponds to an optimal partition of the state space with respect to the K-L divergence rate. The optimal control problem is simplified in an approximate dynamic programming (ADP) framework: First, a relaxation of the policy space is performed, and based on this a parameterization of the set of optimal policies is introduced. This makes possible a stochastic approximation approach to compute the best policy within a given parameterized class. The algorithm can be implemented using a single sample path of the Markov chain. Convergence is established using the ODE method. Examples illustrate the theoretical results, and show remarkably low variance and fast convergence.
Keywords :
Markov processes; cost optimal control; dynamic programming; infinite horizon; reduced order systems; state-space methods; Kullback-Leibler divergence rate; ODE method; aggregating Markov chains; approximate dynamic programming; infinite-horizon average cost optimal control; model reduction; optimal partition; optimal policy; simulation-based framework; simulation-based method; state aggregation; state space; stochastic approximation approach; Application software; Biological system modeling; Computational modeling; Convergence; Cost function; Dynamic programming; Optimal control; Reduced order systems; State-space methods; Stochastic processes;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Decision and Control, 2009 held jointly with the 2009 28th Chinese Control Conference. CDC/CCC 2009. Proceedings of the 48th IEEE Conference on
Conference_Location :
Shanghai
ISSN :
0191-2216
Print_ISBN :
978-1-4244-3871-6
Electronic_ISBN :
0191-2216
Type :
conf
DOI :
10.1109/CDC.2009.5400533
Filename :
5400533
Link To Document :
بازگشت