Multifidelity Approximate Bayesian Computation with Sequential Monte Carlo Parameter Sampling

Multifidelity approximate Bayesian computation (MF-ABC) is a likelihood-free technique for parameter inference that exploits model approximations to significantly increase the speed of ABC algorithms [T. P. Prescott and R. E. Baker, SIAM/ASA J. Uncertain. Quantif., 8 (2020), pp. 114--138]. Previous work has considered MF-ABC only in context rejection sampling, which does not explore space particularly efficiently. In this work, we integrate multifidelity approach with sequential Monte Carlo (ABC-SMC) algorithm into new MF-ABC-SMC algorithm. We show improvements generated by each ABC-SMC efficiency generating samples estimates from posterior are amplified when two techniques used together.