Statistical inverse problems for network dynamics

Summary In many stochastic systems the estimation of the model parameters is complicated by the fact that we do not observe these processes themselves, but rather a mapping or transform of them. One could think of a stochastic process that models the number of molecules of substrates in a biochemical reaction network within a growing cell for which we want to infer how unobserved cellular processes determine the intensity of the counting process. We consider the phenomena that can be modelled as infinite-server queues that are intrinsically more complex than the well-studied M/G/infinity queue. For instance, we consider models with Markov-modulated arrival rates, and models that take into account feedback. Hidden Markov models naturally play a role in the analysis. Of interest is the estimation of the model parameters and asymptotic properties of the estimators. In addition, we are interested in the non-parametric estimation of service time distributions in case of general service times.
Supervisors Mathisca de Gunst (VU), Bartek Knapik (VU), Michel Mandjes (UvA)
PhD Student

Birgit Sollie


VU University Amsterdam (VU)