Importance Sampling for Stochastic Simulations

Importance sampling is one of the classical variance reduction techniques for increasing the efficiency of Monte Carlo algorithms for estimating integrals. The basic idea is to replace the original random mechanism in the simulation by a new one and at the same time modify the function being integrated. In this paper the idea is extended to problems arising in the simulation of stochastic systems. Discrete-time Markov chains, continuous-time Markov chains, and generalized semi-Markov processes are covered. Applications are given to a GI/G/1 queueing problem and response surface estimation. Computation of the theoretical moments arising in importance sampling is discussed and some numerical examples given.