hdp_posterior: Posterior sampling chain across activated DPs.

In nicolaroberts/hdp: Hierarchical Dirichlet Process for categorical count data

Description

Run a Gibbs sampler over the activated DP nodes of a Hierarchichal Dirichlet Process. Each iteration re-assigns the cluster allocation of every data item. Run burnin iterations, and then collect n samples from the chain with space iterations between each collected sample. To plot output, see plot_lik, plot_numcluster, and plot_data_assigned. Can collect multiple independent HDP sampling chains in a hdpSampleMulti object via hdp_multi_chain. Components are extracted via hdp_extract_components.

Usage

hdp_posterior(hdp, burnin, n, space, cpiter = 1, seed = sample(1:10^7, 1),
  verbosity = 0)

Arguments

hdp	A hdpState object
burnin	The number of burn-in iterations.
n	The number of posterior samples to collect.
space	The number of iterations between collected samples.
cpiter	The number of iterations of concentration parameter sampling to perform after each iteration.
seed	The (integer) seed that can be set to reproduce output. Default is a random seed from 1 – 10^7, reported in the output.
verbosity	Verbosity of debugging statements. 0 (least verbose) – 4 (most verbose). 0 highly recommended - only change for debugging small examples.

Value

A hdpSampleChain object with the salient information from each posterior sample. See hdpSampleChain-class

See Also

hdp_multi_chain, hdp_extract_components, cull_posterior_samples, plot_lik, plot_numcluster, plot_data_assigned

Examples

my_hdp <- hdp_init(ppindex=0, cpindex=1, hh=rep(1, 6), alphaa=rep(1, 3), alphab=rep(2, 3))
my_hdp <- hdp_adddp(my_hdp, 2, 1, 2)
my_hdp <- hdp_adddp(my_hdp, 10, c(rep(2, 5), rep(3, 5)), 3)
my_hdp <- hdp_setdata(my_hdp, 4:13, example_data_hdp)
my_hdp <- dp_activate(my_hdp, 1:13, 2)
my_hdp_chain <- hdp_posterior(my_hdp, 100, 100, 10)

nicolaroberts/hdp documentation built on May 23, 2019, 5:09 p.m.