Package index • neuralsbi

Priors

Define the prior distribution over simulator parameters.

priors: Priors for neural simulation-based inference
prior_uniform(): Box-uniform (independent uniform) prior
prior_normal(): Independent normal prior
prior_custom(): Build a prior from arbitrary sampling / density functions
sample_prior(): Draw samples from a prior
within_support(): Test whether parameters lie within the prior support

Simulation and training

Generate training data and fit a neural posterior estimator.

simulate_for_sbi(): Run a simulator over prior draws
npe(): Neural Posterior Estimation (NPE)
npe_sequential(): Sequential NPE with truncated-prior proposals (TSNPE)
density_estimator: Conditional density estimators
embedding_mlp(): Embedding (summary) networks for structured observations

Running the simulator

Parallel execution and progress reporting.

nsbi_parallel parallel-simulation: Running the simulator in parallel
nsbi_progress progress-bars: Progress reporting

Working with the posterior

Condition on data, then sample, evaluate, and summarize.

posterior(): Posterior objects
sample(): Draw samples (S3 generic)
sample(<nsbi_posterior>): Sample from a posterior
sample_posterior(): Sample from a posterior (non-generic alias)
log_prob(): Posterior log-density
map_estimate(): Maximum a posteriori (MAP) estimate
as.data.frame(<nsbi_samples>) summary(<nsbi_samples>) summary(<nsbi_posterior>) summary(<nsbi_npe>): Summaries and tidy accessors

Diagnostics

Calibration and predictive checks for a fitted posterior.

diagnostics: Posterior diagnostics
sbc(): Simulation-Based Calibration (SBC)
expected_coverage(): Expected coverage of central credible intervals
tarp(): TARP expected coverage
c2st(): Classifier two-sample test (C2ST)
posterior_predictive(): Posterior predictive draws

Plotting

pairplot(): Visualize posterior samples
plot_sbc(): Plot an SBC rank histogram
plot_coverage(): Plot nominal vs. empirical credible-interval coverage
plot_tarp(): Plot TARP expected coverage
plot_posterior_predictive(): Plot posterior predictive checks

Benchmark tasks

Reference inference problems shared by the tests and benchmarks.

task_gaussian_linear() task_two_moons() task_slcp() task_sir(): Benchmark tasks