prxgradcholf: Penalized least-square estimation of Cholesky factor

In irenecrsn/covchol: Covariance Cholesky factor estimation

Description

Solve the following optimization problem

\hat{L} = \arg \min_{L} ||LL^t - Sigma||_F* + ||L||_1,off

Usage

prxgradcholf(
  X,
  L = diag(ncol(X)),
  eps = 0.01,
  alpha = 0.5,
  maxIter = 100,
  lambda = 0,
  normalize = TRUE
)

cholpathf(
  X,
  lambdas = NULL,
  L = diag(ncol(X)),
  eps = 1e-08,
  maxIter = 1000,
  normalize = TRUE
)

Arguments

X	data from which to obtain the path
L	initial cholesky factor
eps	convergence threshold for the proximal gradient
alpha	line search rate
maxIter	the maximum number of iterations
lambda	penalization coefficient
normalize	logical
lambdas	increasing sequence of lambdas

Details

cholpath returns the path of regularized estimator on a sequence of lambda parameters

Value

a list with the output of the optimization:

* N * L the estimated L matrix * lambda * diff the value of the last relative decrease * objective the value of the objective function * iter number of iterations

irenecrsn/covchol documentation built on June 1, 2020, 3:31 a.m.