Splitting algorithms for rare event simulation over long time intervals (Q2240483)

The problem of interest is to estimate exit probabilities of a discrete time process \(\{X_i^n\}\) from a bounded open set \(D \subset \mathbb R^d\) over a time interval \([0,T (n)]\). The main focus of this paper is a Monte Carlo estimation of the probability, and in particular on the use of splitting methods in the estimation of exit probabilities. Within the setting of splitting schemes, there are those of interest which allow killing to improve efficiency without introducing bias, with one of the most well known of such schemes being the RESTART method. The authors focus mainly on them. They show that, in contrast to alternatives such as importance sampling, the decay rate of the second moment does not degrade as \(T \rightarrow \infty \). In the course of the analysis, they develop some related large deviation estimates that apply when the time interval of interest depends on the large deviation parameter.