Title: Multifidelity methods for approximate Bayesian computation
Authors: Thomas Prescott - University of Oxford (United Kingdom) [presenting]
Ruth Baker - University of Oxford (United Kingdom)
Abstract: A vital stage in the mathematical modeling of real-world dynamical systems is the calibration of a model's parameters to observed data. Likelihood-free parameter inference methods, such as approximate Bayesian computation (ABC), build Monte Carlo estimates by comparing the experimental data with large numbers of model simulations. However, the computational expense of generating these simulations forms a significant bottleneck in the practical application of such methods. Simulations of corresponding cheap, low-fidelity models can be used to reduce the computational expense of building these samples, at the cost of introducing additional variance to the resulting estimates. The variance costs and computational benefits can optimally balanced, to characterize the optimal choice of how often to simulate from cheap, low-fidelity models in place of expensive, high-fidelity models in Monte Carlo ABC algorithms. The resulting early accept/reject multifidelity ABC algorithm is shown to give improved performance over high-fidelity approaches in the context of both importance sampling and sequential Monte Carlo sampling.