run_gibbsflow_ais: Run Gibbs flow annealed importance sampler

In jeremyhengjm/GibbsFlow: Gibbs Flow for Approximate Transport

Description

Run Gibbs flow annealed importance sampler

Usage

run_gibbsflow_ais(
  prior,
  likelihood,
  nparticles,
  timegrid,
  lambda,
  derivative_lambda,
  compute_gibbsflow,
  mcmc,
  gibbsvelocity = NULL
)

Arguments

prior	list with keys: logdensity evaluates log prior density, gradlogdensity returns its gradient, rinit samples from the prior distribution
likelihood	list with keys: logdensity samples from proposal, gradlogdensity returns its gradient
nparticles	number of particles
timegrid	vector describing numerical integration times
lambda	vector describing tempering schedule
derivative_lambda	time derivative of tempering schedule
compute_gibbsflow	function computing Gibbs flow
mcmc	list with keys: choice specifies type of MCMC method, parameters specifies algorithmic tuning parameters, nmoves specifies number of MCMC move per temperature

Value

list with keys: xtrajectory trajectories, xparticles particles at terminal time, ess effective sample size, log_normconst log normalizing constant, acceptprob MCMC acceptance probabilities

See Also

run_gibbsflow_smc if resampling is desired

jeremyhengjm/GibbsFlow documentation built on Feb. 14, 2021, 9:21 p.m.