Non-asymptotic analysis of ensemble Kalman updates: effective dimension and localization
Non-asymptotic analysis of ensemble Kalman updates: effective dimension and localization
Abstract Many modern algorithms for inverse problems and data assimilation rely on ensemble Kalman updates to blend prior predictions with observed data. Ensemble Kalman methods often perform well with a small ensemble size, which is essential in applications where generating each particle is costly. This paper develops a non-asymptotic analysis …