parallel_h_fusion_TA_mixG: Time-adapting Hierarchical Monte Carlo Fusion

In rchan26/mixGaussTempering: What the Package Does (One Line, Title Case)

Description

Hierarchical Time-adapting Monte Carlo Fusion with base level with nodes that are tempered mixture Gaussians with same weights, means, sds components

Usage

parallel_h_fusion_TA_mixG(N_schedule, global_T, start_beta, C_schedule, L,
  base_samples, weights, means, sds, study = F)

Arguments

N_schedule	vector of length (L-1), where N_schedule[l] is the number of samples per node at level l
global_T	time T for time-adapting fusion algorithm
start_beta	beta for the base level
C_schedule	vector of length (L-1), where C_schedule[l] is the number of samples to fuse for level l
L	total number of levels in the hierarchy
base_samples	list of length (1/start_beta), where samples_to_fuse[c] containg the samples for the c-th node in the level
weights	vector: weights of mixture Gaussian
means	vector: means of mixture Gassuan
sds	vector: st.devs of mixture Gaussian
study	boolean value: defaults to F, determines whether or not to return acceptance probabilities

Value

samples samples from hierarchical fusion

time vector of length (L-1), where time[l] is the run time for level l

rho_acc vector of length (L-1), where rho_acc[l] is the acceptance rate for first fusion step for level l

Q_acc vector of length (L-1), where Q_acc[l] is the acceptance rate for second fusion step for level l

input_betas list of length (L), where input_betas[[l]] is the input betas for level l

output_beta vector of length(L-1), where output_beta[l] is the beta for level l

diffusion_times list of length (L-1), where diffusion_times[[l]] are the scaled/adapted times for fusion in level l

Examples

# setting variables
w_example <- c(0.35, 0.65)
m_example <- c(-3, 8)
s_example <- c(1, 1.5)
b_example <- 1/4

# sampling from tempered density
nsamples <- 500000
base <- hmc_base_sampler_mixG(w_example, m_example, s_example, b_example, nsamples, 4)

test <- parallel_h_fusion_TA_mixG(N_schedule = rep(10000, 2), global_T = 1,
                                  start_beta = b_example, C_schedule = rep(2, 2), L = 3, base_samples = base,
                                  weights = w_example, means = m_example, sds = s_example, study = T)

# plot results
plot_levels_hier_mixG(test, weights = w_example, means = m_example, sds = s_example,
                      from = -15, to = 20, plot_rows = 3, plot_columns = 1, bw = c(0.2, 0.3, 0.4))

rchan26/mixGaussTempering documentation built on June 14, 2019, 3:26 p.m.