get_mh_kernel: Get random walk Metropolis-Hastings kernels

In pierrejacob/debiasedmcmc: Unbiased MCMC with couplings

Get random walk Metropolis-Hastings kernels

Description

This function takes a target (specified through its log-pdf) and a covariance matrix for a Normal random walk proposal, and returns a list containing the keys single_kernel, coupled_kernel corresponding to marginal and coupled MH kernels.

The coupling is done by reflection-maximal coupling of the proposals, and common uniform variable for the accept/reject step. For reflection-maximal couplings, see rnorm_reflectionmax and rmvnorm_reflectionmax.

The returned kernels can then be used in the functions sample_meetingtime or sample_coupled_chains or sample_unbiasedestimator.

Usage

get_mh_kernels(target, Sigma_proposal)

Arguments

target	function taking a vector as input and returning target log-density evaluation
Sigma_proposal	covariance of the Normal random walk proposal

Value

A list containing the keys single_kernel, coupled_kernel.

pierrejacob/debiasedmcmc documentation built on Aug. 22, 2022, 12:41 a.m.