Neural Posterior Estimation with Differentiable Simulators

TL;DR

This paper introduces a neural posterior estimation method that leverages differentiable simulators to improve sample efficiency and posterior accuracy in simulation-based inference tasks.

Contribution

The paper presents a novel approach to neural posterior estimation that utilizes gradients from differentiable simulators, enhancing efficiency and posterior quality.

Findings

Gradient information constrains posterior shape effectively

Method improves sample efficiency in neural posterior estimation

Differentiable simulators enhance Bayesian inference accuracy

Abstract

Simulation-Based Inference (SBI) is a promising Bayesian inference framework that alleviates the need for analytic likelihoods to estimate posterior distributions. Recent advances using neural density estimators in SBI algorithms have demonstrated the ability to achieve high-fidelity posteriors, at the expense of a large number of simulations ; which makes their application potentially very time-consuming when using complex physical simulations. In this work we focus on boosting the sample-efficiency of posterior density estimation using the gradients of the simulator. We present a new method to perform Neural Posterior Estimation (NPE) with a differentiable simulator. We demonstrate how gradient information helps constrain the shape of the posterior and improves sample-efficiency.