Moment closure based parameter inference of stochastic kinetic models

Abstract

Parameter inference for stochastic kinetic models is a topic that spans many disciplines. Although it is possible to carry out exact inference using partial observations of a stochastic process, it is often computationally impractical. In this paper we use the moment closure approximation of the underlying stochastic process as a fast approximation of the likelihood. We show that this approximation is fast and accurate, even when the population numbers are small.