Monte Carlo perturbation algorithms exploit new computer architectures

Fundamental aspects of correlated sampling and differential operator procedures applied to systems of linear equations modelling Markov processes are investigated. Algorithms providing sensitivities (gradients, Jacobians) and perturbation estimates obtained a single simulation experiment are described in detail. In this context it becomes evident that the algorithm for calculating the Jacobians may benefit considerably from the use of vector processors. Mathematical proofs are provided which show that under most conditions a finite relative variance can be obtained for arbitrarily small parameter variations. (orig.)