Rare Events

In a reachability probability problem, one wants to compute the probability p to reach a final state s+, from the initial state s0 of a given model. When the size of the model becomes too large for numerical analysis and when the probability of interest p is too small for direct statistical simulations, rare event acceleration methods become necessary. A possible approach to tackle this problem is to build an importance sampling scheme for which a statistical simulation becomes possible. In order to be useful, this importance sampling has to be built carefully to ensure a reduction of the variance. In [16,17] a method to produce such an importance sampling is described. This approach has been implemented in Cosmos and have been successfully applied to waiting queue example as well as a biological case study [9,12].