Importance sampling
From PBDN
Introduction
Importance sampling is one of the four variance reduction techniques used in the project. The idea behind importance sampling is to choose chemical reactions according to another probability measure compared to that given by the original model. To simulate the system of interest, such a change must be followed by a transformation back to the first model. In this way, quantities of interest (such as mean values) are simulated. However, variance does not remain unchanged. With an appropriate choice of probability function for the chemical reactions, variance can be decreased. Thus, fewer trajectories are needed to reach a satisfactorily smooth PDF describing the system.
Results
There is no direct way to find the optimal importance function. In fact, many functions increase variance and the behaviour of the solution is deteriorated. All functions investigated have had an overall variance increasing effect.
Back to Simulation of Biochemical Reaction Networks main page.