Likelihood-free inference with emulator networks

TL;DR

This paper introduces a novel likelihood-free inference method using probabilistic neural emulator networks within ABC, enabling efficient and accurate Bayesian inference for complex simulation-based models without requiring explicit likelihood functions.

Contribution

The authors develop a new ABC approach employing neural emulator networks to learn synthetic likelihoods, improving inference efficiency and accuracy in high-dimensional, simulation-based models.

Findings

Emulators enable accurate inference on synthetic data.

Method performs well on high-dimensional biophysical neuron models.

Approach eliminates the need for rejection thresholds or distance functions.

Abstract

Approximate Bayesian Computation (ABC) provides methods for Bayesian inference in simulation-based stochastic models which do not permit tractable likelihoods. We present a new ABC method which uses probabilistic neural emulator networks to learn synthetic likelihoods on simulated data -- both local emulators which approximate the likelihood for specific observed data, as well as global ones which are applicable to a range of data. Simulations are chosen adaptively using an acquisition function which takes into account uncertainty about either the posterior distribution of interest, or the parameters of the emulator. Our approach does not rely on user-defined rejection thresholds or distance functions. We illustrate inference with emulator networks on synthetic examples and on a biophysical neuron model, and show that emulators allow accurate and efficient inference even on high-dimensional problems which are challenging for conventional ABC approaches.