compute_loading_elbo: Loading ELBO

In jpmeagher/vbar: Variational Bayes Ancestral Reconstruction

Loading ELBO

Description

Compute the contribution of the loading matrix Evidence Lower Bound (ELBO) of a PLVM given the approximate posterior distribution for loadings, the prior correlation matrix, and ARD precision hyperparameters.

Usage

compute_loading_elbo(
  loading_expectation,
  loading_row_covariance,
  ard_precision,
  loading_prior_correlation_log_det = NULL,
  inv_loading_prior_correlation = NULL,
  loading_prior_correlation = NULL,
  perform_checks = TRUE
)

Arguments

loading_expectation	A D'xL matrix of real numbers, The expected loading matrix.
loading_row_covariance	A LxLxD' array. The covariance of each row of the loading matrix under the approximate posterior distribution.
ard_precision	An L-dimensional vector of real numbers. The ARD precision hyper-parameters on the columns of the loading matrix.
loading_prior_correlation_log_det	A real valued scalar. The log determinant of the loading prior correlation matrix.
inv_loading_prior_correlation	A D'xD' matrix. The inverse of the loading prior correlation matrix.
loading_prior_correlation	A D'xD' matrix. The loading prior correlation matrix.
perform_checks	Logical. Check if function inputs are specified correctly.

Value

A real valued scalar. The contribution of the loading matrix to the ELBO under the approximate posterior distribution.

jpmeagher/vbar documentation built on Nov. 22, 2022, 5:48 a.m.