CMStatistics 2018: Start Registration
View Submission - CMStatistics
B0314
Title: Stein Points Authors:  Francois-Xavier Briol - Imperial College London (United Kingdom) [presenting]
Abstract: An important task in computational statistics is to approximate a posterior distribution with an empirical measure supported on a set of representative points. This talk will focus on methods where the selection of points is essentially deterministic, with an emphasis on achieving accurate approximation when n is small. To this end, we will present a new algorithm called 'Stein Points'. The idea is to exploit either a greedy or a conditional gradient method to iteratively minimise a kernel Stein discrepancy between the empirical measure and the target. Our empirical results demonstrate that Stein Points enable accurate approximation of the posterior at modest computational cost. In addition, theoretical results are provided to establish convergence of the method.