1er Symposium canadien en analyse numérique et calcul scientifique
Simulating State-Dependent Diffusions
Tupper, Paul
Department of Mathematics, Simon Fraser University

We propose a framework for modelling stochastic systems with variable diffusion rates which satisfy detailed balance (or in other terminology, time-reversibility). Rather than specifying the dynamics through a state-dependent drift and diffusion coefficients, we specify an equilibrium probability density and a state-dependent diffusion coefficient. We argue that our framework is more natural from the modelling point of view and has a distinct advantage in situations where either the equilibrium probability density or the diffusion coefficient is discontinuous. We introduce a numerical method for simulating dynamics in our framework that samples from the equilibrium probability density exactly and elegantly handles discontinuities in the coefficients. This is joint work with Xin Yang.

Mardi, 18 juin, 17h30
Salle Des Plaines C