Multilevel bootstrap particle filter

We consider situations where the applicability of sequential Monte Carlo particle filters is compromised due to expensive evaluation weights. To alleviate this problem, we propose a new filter algorithm based on multilevel approach. show that resulting bootstrap (MLBPF) retains strong law large numbers as well central limit theorem classical under mild conditions. Our numerical experiments demonstrate up 85% reduction in computation time compared filter, certain settings. While it should be acknowledged highly application dependent, and similar gain not expected for all applications across board, believe substantial improvement settings makes MLBPF an important addition family methods.