Approximate inference for stochastic reaction processes.
Andreas Ruttor, Guido Sanguinetti and Manfred Opper
Learning and Inference in Computational Systems Biology
We discuss the problem of statistical inference for Markov jump
processes modelling biochemical reactions. Using a variational
formulation of exact inference we derive two different approximations.
A weak noise approach within a diffusion approximation is relevant
when the number of individuals of a given species is rather large. On
the other hand a mean field approximation takes the discreteness of
the number of individuals into account but neglects correlations.