AI Seminar: Continuous-discrete smoothing of diffusions


Frank van der MeulenAssociate Professor in Electrical Engineering, Mathematics and Computer Science at Delft University of Technology, the Netherlands.


In this talk I will consider sampling unobserved parts of the path of a discretely observed diffusion generated by a stochastic differential equation. I will explain how this can be done within the framework of guided proposals. It turns out that the problem can be dealt with while avoiding linearisation of the drift and diffusion coefficients. In this sense the proposed method is a refinement of Kalman smoothing. The method is illustrated on a landmark matching problem.