Search Results

  1. P. Lassila and J. Virtamo, Efficient Monte Carlo Simulation of Product Form Systems, in Fourteenth Nordic Teletraffic Seminar, NTS-14, pp. 355-366, 1998, Copenhagen, Denmark (pdf)(bib)
    Abstract: We discuss efficient Monte Carlo simulation techniques for estimating performance measures of product form systems, e.g. the multiservice loss system. First, two methods for generating the samples are discussed, the traditional Monte Carlo and the so called Gibbs sampler. Then a variance reduction method, the conditional expectation method, is presented where the idea is to utilize known analytical results to the maximum degree. This is possible for systems with a product form probability distribution for which conditional one-dimensional expectations can easily be precomputed. The method is independent of the way the samples are generated. Finally, we study the use of heuristic importance sampling distribution in connection with the traditional Monte Carlo method and the conditional expectation method.