Abstract:
|
In simulation-based inferences for partially observed Markov process models, the by-product of the Monte Carlo filtering is an approximation of the likelihood. Consequently, different stochastic optimization algorithm can be applied to estimate the parameters of the underlying models. In this paper, we view iterated filtering as a pseudo-proximal map of the likelihood function. As proximal map is an efficient optimization approach in the optimization literature, we show that stochastic proximal map also inherit that high convergence rate while the algorithm is quite simple as the step size cancel out in an iterated filtering framework, only the mean of each perturbation filter left. We show that this new algorithm is favourably compared to other previous iterated filtering in a toy example and a challenging scientific problem of modeling infectious diseases.
|