Book chapter
Accelerated Convergence through Extrapolation for the Discounted Return in Markov Chains
Numerical Solution of Markov Chains, pp 667-669
1991
Abstract
We are concerned with methods for accelerating the convergence of iterative methods for computing the discounted return for infinite-horizon, finite-state Markov chains. We define the set of states as S = {1,2,…,n} transition matrix P = {pij
}, cost vector c = {c
i}, and discount factor α ∈ (0,1). The objective is to determine, for each state i, the long-range expected discounted cost. This cost, f
i, is determined by
f
i
=
α
∑
j
∈
S
p
i
j
f
j
+
c
i
,
i
∈
S
. The existence of a unique f is guaranteed.
Metrics
4 Record Views
Details
- Title
- Accelerated Convergence through Extrapolation for the Discounted Return in Markov Chains
- Creators
- Jeffrey L Popyack - Drexel University
- Contributors
- William J. Stewart (Editor)
- Publication Details
- Numerical Solution of Markov Chains, pp 667-669
- Publisher
- CRC Press
- Edition
- 1
- Resource Type
- Book chapter
- Language
- English
- Academic Unit
- College of Computing and Informatics
- Other Identifier
- 991021870312304721