Mean-field limit and numerical analysis for Ensemble Kalman Inversion: linear setting

Zhiyan Ding, Qin Li

Published 2019-08-15Version 1

Ensemble Kalman inversion (EKI) is a method introduced in [14] to find samples from the target posterior distribution in the Bayesian formulation. As a deviation from Ensemble Kalman filter [6], it introduces a pseudo-time along which the particles sampled from the prior distribution are pushed to fit the profile of the posterior distribution. To today, however, the thorough analysis on EKI is still unavailable. In this article, we analyze the continuous version of EKI, a coupled SDE system, and prove the solution to this SDE system convergences, as the number of particles goes to infinity, to the target posterior distribution in Wasserstein distance in finite time.