Fast simulation of rare events using reverse-time models

Abstract

We consider the estimation by simulation of probabilities of rare but potentially damaging events in Markovian systems. The problem is that the rarity of the event mitigates against its frequent occurrence in simulation. Here we show how a reverse-time model may be simulated in place of the forward-time system in order to provide significantly faster estimates of the rare probabilities. An example of an ALOHA system will be used and simulation speedup compared with alternative methods based on Large Deviations theory.