Abstract : This paper aims at solving the problem of parameter identification in interaction networks. Purely discontinuous Markov processes are the most appropriate choice for modeling such systems, and it is of paramount importance to estimate the unknown characteristics of the model given the measured data. The model induces a Fokker-Planck-Kolmogorov equation along with moment equations, and achieving parametric identification based on direct solutions of these equations has remained elusive. We propose a novel approach which utilizes applied stochastic analysis for continuous-time Markov processes to lift the curse of dimensionality in parametric identification. We illustrate through two case studies of biological network modeling.
