Title :
Single sample path-based sensitivity analysis of Markov processes using uniformization
Author :
Liu, Zikuan ; Tu, Fengsheng
Author_Institution :
Dept. of Appl. Math., Hebei Univ. of Technol., Tianjin, China
fDate :
4/1/1999 12:00:00 AM
Abstract :
Using the notion of perturbation realization factor, Cao and Chen (1997) provide sensitivity formulas of discrete-time Markov chains and uniformizable Markov processes. In this paper, the estimators of the realization factors of a uniformizable Markov process are provided by its uniformized Markov chain. It is proved that estimators given by the uniformized Markov chain have smaller variances than those provided by the original Markov process, which are increasing functions of the uniformization parameter and thus have minimal solutions
Keywords :
Markov processes; matrix algebra; perturbation techniques; probability; sensitivity analysis; Markov chains; Markov processes; perturbation; probability matrix; sensitivity analysis; single sample path; uniformization; Analytical models; Markov processes; Mathematics; Particle measurements; Performance analysis; Sensitivity analysis; State estimation; State-space methods; Stochastic processes; Uncertainty;
Journal_Title :
Automatic Control, IEEE Transactions on