golazo: Performs Graphical Oriented LAZy Optimization by optimizing...

In pzwiernik/mtp2: Working with Gaussian totally positive distributions

Description

This function implements a simple block-coordinate descent algorithm to find the maximum of the regularized Gaussiann log-likelihood with a an assymetric penalty of lasso type.

Usage

golazo(
  S,
  rho = NULL,
  L = NULL,
  U = NULL,
  tol = 1e-07,
  pos.constr = TRUE,
  output = TRUE
)

Arguments

S	the sample covariance matrix
rho	positive penalty (can be Inf)
L	matrix of lower penalties (can be -Inf)
U	matrix of upper penalties (can be Inf)
tol	the convergence tolerance (default tol=1e-8)
pos.constr	if TRUE (default) penalizes positive K_ij, if FALSE performs the standard dual graphical lasso.
output	if TRUE (default) the output will be printed.

Value

the optimal value of the concentration matrix

the number of iterations the algorithm needed to converge

the corresponding value of the log-likelihood

Examples

print(TRUE)

pzwiernik/mtp2 documentation built on Aug. 9, 2020, 12:34 p.m.