Estimating small probabilities exactly

One problem in rare event estimation is to establish how likely it is for a Markov process (eg the solution path to an SDE) to hit some set of interest before another much more likely set. If the process can be simulated at least approximately, there are computationally efficient estimators of these probabilities, but they are typically biased in a way which is difficult to quantify. This talk is about simulating these rare events in multiple dimensions without approximation, so the resulting estimator is unbiased.