marginal_p_gaussian: Marginal Probabilities for Gaussian distributed data

In jpkrooney/rcorex: An Implementation of Total Correlation Explanation

Marginal Probabilities for Gaussian distributed data

Description

Internal function to estimate the marginals for Gaussian distributed data.

Usage

marginal_p_gaussian(x_i, thetai)

Arguments

x_i	A single variable/column of data
thetai	Estimated parameters corresponding to the single variable/column of data provided to x_i

Details

Estimate the marginals for a single column of data x_i for each hidden variable with dimension dim_hidden. For continuous data we estimate p≤ft(y_{j} \mid x_{i}\right) / p≤ft(y_{j}\right) indirectly by using Bayes rule to rewrite this as p≤ft(x_{i} \mid y_{j}\right) / p≤ft(x_{i}\right), which can be estimated.

Value

A three dimensional array of marginals with dimensions: (n_hidden, dim_hidden, n_samples) - i.e. marginals for each data point in x_i given current n_hidden x dim_hidden parameter estimates

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