Abstract:Models of stochastic processes are widely used in almost all fields of science. Theory validation, parameter estimation, and prediction all require model calibration and statistical inference using data. However, data are almost always incomplete observations of reality. This leads to a great challenge for statistical inference because the likelihood function will be intractable for almost all partially observed stochastic processes. This renders many statistical methods, especially within a Bayesian framework, impossible to implement. Therefore, computationally expensive likelihood-free approaches are applied that replace likelihood evaluations with realisations of the model and observation process. For accurate inference, however, likelihood-free techniques may require millions of expensive stochastic simulations. To address this challenge, we develop a new method based on recent advances in multilevel and multifidelity. Our approach combines the multilevel Monte Carlo telescoping summation, applied to a sequence of approximate Bayesian posterior targets, with a multifidelity rejection sampler to minimise the number of computationally expensive exact simulations required for accurate inference. We present the derivation of our new algorithm for likelihood-free Bayesian inference, discuss practical implementation details, and demonstrate substantial performance improvements. Using examples from systems biology, we demonstrate improvements of more than two orders of magnitude over standard rejection sampling techniques. Our approach is generally applicable to accelerate other sampling schemes, such as sequential Monte Carlo, to enable feasible Bayesian analysis for realistic practical applications in physics, chemistry, biology, epidemiology, ecology and economics.

A Stratified Simulation Scheme for Inference in Bayesian Belief Networks

Low-Budget Simulation-Based Inference with Bayesian Neural Networks

All-in-one simulation-based inference

Fast, accurate and lightweight sequential simulation-based inference using Gaussian locally linear mappings

Stratified sampling and bootstrapping for approximate Bayesian computation

Pseudo-Likelihood Inference

Backward Simulation in Bayesian Networks

Efficient Multifidelity Likelihood-Free Bayesian Inference with Adaptive Computational Resource Allocation

Contrastive Neural Ratio Estimation for Simulation-based Inference

Optimal simulation-based Bayesian decisions

Compositional simulation-based inference for time series

A simulated annealing algorithm for randomizing weighted networks

Amortized Simulation-Based Frequentist Inference for Tractable and Intractable Likelihoods

Amortized Bayesian Decision Making for simulation-based models

Advancing Tools for Simulation-Based Inference

Probabilistic Surrogate Networks for Simulators with Unbounded Randomness

Neural Methods for Amortized Inference

Inference from Iterative Simulation Using Multiple Sequences

Multifidelity multilevel Monte Carlo to accelerate approximate Bayesian parameter inference for partially observed stochastic processes

Preconditioned Neural Posterior Estimation for Likelihood-free Inference

Belief Updating by Enumerating High-Probability Independence-Based Assignments