Arbitrary Marginal Neural Ratio Estimation for Simulation-based Inference
This addresses the problem of interpreting posteriors in simulation-based inference for scientists, though it appears incremental as it builds on existing neural ratio estimation techniques.
The paper tackles the challenge of performing Bayesian inference with complex stochastic simulators by introducing a method for amortized inference over arbitrary parameter subsets without numerical integration, demonstrating its applicability on binary black hole parameter inference from gravitational wave observations.
In many areas of science, complex phenomena are modeled by stochastic parametric simulators, often featuring high-dimensional parameter spaces and intractable likelihoods. In this context, performing Bayesian inference can be challenging. In this work, we present a novel method that enables amortized inference over arbitrary subsets of the parameters, without resorting to numerical integration, which makes interpretation of the posterior more convenient. Our method is efficient and can be implemented with arbitrary neural network architectures. We demonstrate the applicability of the method on parameter inference of binary black hole systems from gravitational waves observations.