Thursday 14 Dec 2017: Statistical Science seminar-Modelling discontinuities using Voronoi tessellations
Prof J.P. Gosling - University of Leeds
The standard Gaussian process emulator (or kriging estimator) has a drawback in that the computer code output (or spatial field) is assumed to be homogeneous. Computer codes can behave very differently in various regions of the input space, and we will show the issues that can occur when blindly applying emulator methods. To counteract this issue, we introduce a piecewise Gaussian process model to deal with this problem where the input space is divided into separate regions using Voronoi tessellations (also known as Dirichlet tessellations, Thiessen polygons or the dual of the Delaunay triangulation). We demonstrate our method’s utility with an application to cloud modelling and discuss the problem of visualising model regime changes in multidimensional space.