This paper discusses the application of moment closures to continuous Markov chains derived from process algebras such as GPEPA and MASSPA. Two related approaches are being investigated. Firstly we re-formulate normal moment closure in a process algebra framework using Isserlis' theorem. Secondly we apply a mixture of this normal closure and less precise moment closures for the purpose of reducing coupling between ordinary differential equations (ODE) derived from the underlying Markov chain. We present three case-studies to show how both normal and inhomogeneous moment closures can significantly improve the numerical accuracy of ODE moment approximations.
Information from pubs.doc.ic.ac.uk/moment-closures.