Papers

Estimating Marginal Likelihoods in Likelihood-Free Inference via Neural Density Estimation

P Bastide, A Estoup, JM Marin, J Stoehr - arXiv preprint arXiv:2507.08734, 2025 - arxiv.org
Statistics paper stat.CO Suggest

… Crucially, the method is not specific to simulation-based inference and could be applied in any Bayesian setting where a posterior sample is available and a surrogate …

Link to paper

BibTeX

@article{2507.08734v1,
Author        = {Paul Bastide and Arnaud Estoup and Jean-Michel Marin and Julien Stoehr},
Title         = {Estimating Marginal Likelihoods in Likelihood-Free Inference via Neural
  Density Estimation},
Eprint        = {2507.08734v1},
ArchivePrefix = {arXiv},
PrimaryClass  = {stat.CO},
Abstract      = {The marginal likelihood, or evidence, plays a central role in Bayesian model
selection, yet remains notoriously challenging to compute in likelihood-free
settings. While Simulation-Based Inference (SBI) techniques such as Sequential
Neural Likelihood Estimation (SNLE) offer powerful tools to approximate
posteriors using neural density estimators, they typically do not provide
estimates of the evidence. In this technical report presented at BayesComp
2025, we present a simple and general methodology to estimate the marginal
likelihood using the output of SNLE.},
Year          = {2025},
Month         = {Jul},
Url           = {http://arxiv.org/abs/2507.08734v1},
File          = {2507.08734v1.pdf}
}

Share