Low variance couplings for stochastic models of intracellular processes with time-dependent rate functions
For researchers in stochastic modeling of intracellular processes, this method reduces variance in path comparisons, enabling more efficient sensitivity analysis and expectation estimation.
The paper introduces stacked coupling for stochastic biochemical models with time-dependent parameters, achieving exceptionally low variance between generated paths, which improves parametric sensitivity computation and multilevel Monte Carlo estimation.
A number of coupling strategies are presented for stochastically modeled biochemical processes with time-dependent parameters. In particular, the stacked coupling is introduced and is shown via a number of examples to provide an exceptionally low variance between the generated paths. This coupling will be useful in the numerical computation of parametric sensitivities and the fast estimation of expectations via multilevel Monte Carlo methods. We provide the requisite estimators in both cases.