update_tuning_mv_mat: this function updates multiple block Gaussian random walk...

In jtipton25/BayesMRA: Bayesian Multi-Resolution Gaussian Process Approximations

this function updates multiple block Gaussian random walk proposal Metropolis-Hastings tuning parameters

Description

this function updates multiple block Gaussian random walk proposal Metropolis-Hastings tuning parameters

Usage

update_tuning_mv_mat(
  k,
  accept,
  lambda,
  batch_samples,
  Sigma_tune,
  Sigma_tune_chol
)

Arguments

k	is a positive integer that is the current MCMC iteration.
accept	is a p-dimensional vector of positive number between 0 and 1 that represents the batch accpetance rate. The target of this adaptive tuning algorithm is an acceptance rate of between 0.44 (a univariate update) and 0.234 (5+ dimensional upate).
lambda	is a is a p-dimensional vector of positive scalars that scales the covariance matrix Sigma_tune and diminishes towards 0 as k increases.
batch_samples	is a 50 \times d \times p dimensional array that consists of the 50 batch samples of the d-dimensional for each of the pp parameter groups being sampled.
Sigma_tune	is a d \times d \times p array of d \times d positive definite covariance matrix of the batch samples used to generate the multivariate Gaussian random walk proposal.
Sigma_tune_chol	is the d \times d \times p array of d \times d Cholesky decompositions of Sigma_tune.

jtipton25/BayesMRA documentation built on Feb. 28, 2024, 1:27 p.m.