Infinitesimal perturbation analysis is a sensitivity-analytic technique for discrete event dynamic systems using single sample path performance measures. The basic approach is to obtain unbiased or strongly consistent estimates for derivatives of the stationary performance measures. For infinitesimal perturbation analysis of Markov chains, Cao, Yuan and Qiu [1996] proposed a novel approach in terms of the realization factor. Perturbation analysis of Markov chains is quite extensive, see Chapter 11 in Li's book: Constructive Computation in Stochastic Models with Applications: RG-Factorizations. On the other hand, He also used the perturbation analysis of Markov chains to deal with ESS of stochastic evolutionary games.
넶0
Infinitesimal perturbation analysis is a sensitivity-analytic technique for discrete event dynamic systems using single sample path performance measures. The basic approach is to obtain unbiased or strongly consistent estimates for derivatives of the stationary performance measures. For infinitesimal perturbation analysis of Markov chains, Cao, Yuan and Qiu [1996] proposed a novel approach in terms of the realization factor. Perturbation analysis of Markov chains is quite extensive.
2021-09-29 01:00