Glarmadillo (version 1.1.1)

glarma: Solve Graphical Lasso with Armadillo

Description

This function solves the Graphical Lasso (GLasso) problem using the Armadillo library. GLasso is a technique used in statistical learning and network analysis to estimate sparse inverse covariance matrices from observed data.

Usage

glarma(s, rho, mtol = 1e-04, maxIterations = 10000, ltol = 1e-06)

Value

Returns a covariance matrix W and a estimated sparse inverse covariance matrix Theta estimated by solving the Graphical Lasso problem. The sparsity is controlled by the 'rho' parameter.

Arguments

s

A symmetric, positive-definite sample covariance matrix. It should be a square matrix representing the covariance matrix of the variables.

rho

A positive scalar representing the regularization parameter. It controls the sparsity level of the inverse covariance matrix.

mtol

A numeric value representing the convergence threshold for the main algorithm. It determines the condition under which the iterative process will stop. Default is 1e-4.

maxIterations

An integer value specifying the maximum number of iterations allowed for the algorithm. Default is 10000.

ltol

A numeric value representing the convergence threshold for the Lasso solver. It is used to control the Lasso solving process within the algorithm. Default is 1e-6.

Examples

# Generate a sample covariance matrix
s <- matrix(runif(100), nrow = 10)
s <- t(s) %*% s
# Solve the Graphical Lasso problem with default parameters
inv_cov_matrix <- glarma(s, rho = 0.1)
# Solve with custom convergence thresholds and maximum iterations
inv_cov_matrix <- glarma(s, rho = 0.1, mtol = 1e-5, maxIterations = 5000, ltol = 1e-6)