calculate_marginals_on_samples: Calculate marginals on data

In jpkrooney/rcorex: An Implementation of Total Correlation Explanation

Calculate marginals on data

Description

Internal function to calculate marginals given data and parameter estimates

Usage

calculate_marginals_on_samples(
  data,
  theta,
  marginal_description,
  log_p_y,
  dim_visible = NULL,
  returnratio = TRUE,
  logpx_method
)

Arguments

data	Data provided by user
theta	List of estimated parameters
marginal_description	Character string which determines the marginal distribution of the data. A single marginal description applies to all variables in biocorex.
log_p_y	A 2D matrix representing the log of the marginal probability of the latent variables
dim_visible	The dimension of the data provided - i.e. the number of discrete levels that exist in the data. Must be positive integer.
returnratio	A Boolean that returns log_marg_x_4d for the value TRUE and log_p_xi_given_y_4d for the value FALSE. Intended for development use and may not be retained long term.
logpx_method	EXPERIMENTAL - A character string that controls the method used to calculate log_p_xi. "pycorex" uses the same method as the Python version of biocorex, "mean" calculates an estimate of log_p_xi by averaging across n_hidden estimates.

Details

Calculate the value of the marginal distribution for each variable, for each dimension of each hidden variable and each sample

Value

4D array of dimensions: (n_hidden, n_samples, n_visible, dim_hidden ) where n_samples is the number of rows in the user provided data, and n_visible is the number of columns. Returned data is result fo the calculation log ≤ft( \frac{p≤ft(y_{j} \mid x_{i}\right)}{p≤ft(y_{j}\right)} \right)) for each j,sample,i,y_j

jpkrooney/rcorex documentation built on July 25, 2022, 1:37 a.m.