neuralsbi: Neural Simulation-Based Inference

A native R implementation of neural simulation-based inference (SBI) methods. Given a prior over parameters and a simulator, 'neuralsbi' trains a conditional neural density estimator for likelihood-free Bayesian inference: Neural Posterior Estimation approximates the posterior directly and samples it in a forward pass, Neural Likelihood Estimation learns a surrogate likelihood that is sampled with MCMC, handles repeated independent observations, and can be exported as 'Stan' code for use inside a larger model, and Neural Ratio Estimation trains a classifier for the likelihood-to-evidence ratio instead of either density. Neural estimators run on the 'torch' back end. This package is developed for applied researchers who want an R-native approachable SBI interface with sensible defaults and built-in posterior diagnostics.

See also

Useful links:

• https://neuralsbi.pedrodelima.com/

• https://github.com/pedroliman/neuralsbi

• Report bugs at https://github.com/pedroliman/neuralsbi/issues

Author

Maintainer: Pedro Nascimento de Lima plima@rand.org (ORCID)

Authors:

• Pedro Nascimento de Lima plima@rand.org (ORCID)