Contents of this page:
An old simulation technique called the 'splitting method' has become
a popular competitor to the Importance Sampling method for rare event
The method is based on the splitting of every successful simulation path
into multiple paths with smaller weights, and doing this recursively.
We analyze the method for asymptotic properties and parameter choices for optimal efficient implementations. Furthermore, several new techniques and enhancements are proposed and similarly analyzed, using actual simulation results. The results are promising in the sense that for a class of non-trivial systems the Importance Sampling efficiency can be achieved robustly, so without e.g. the IS drawbacks of tilt-factor estimation.
To download my thesis `The splitting method in rare event simulation', which I finished in 2000, press: thesis.
|ASM Lithography BV
De Run 1110
5503 LA Veldhoven