Guido Sanguinetti abstract
Title: Cox process representation and inference for stochastic reaction-diffusion processes
Abstract: Spatio-temporal models play a pivotal role in our understanding of a variety of natural phenomena in many disciplines. Stochastic reaction-diffusion processes are an increasingly important paradigm for distributed dynamical systems, yet they are notoriously difficult to simulate and calibrate to observational data. Here we show that stochastic reaction-diffusion processes can be accurately approximated by spatio-temporal Cox processes. This novel connection arises naturally as an approximation of the Poisson representation of a spatial description of stochastic chemical reaction systems, giving both novel conceptual insights and allowing us to leverage a large statistical literature for efficient inference techniques. We demonstrate the power of this approach by applying it to several synthetic and real data examples from a variety of application domains.
Our results demonstrate that the method is highly accurate and computationally efficient, providing a valuable tool for parameter inference in a wide range of scientific disciplines.
Joint work with David Schnoerr and Ramon Grima