fista.LpS: A function to solve low rank plus sparse model estimation...

In LSVAR: Estimation of Low Rank Plus Sparse Structured Vector Auto-Regressive (VAR) Model

Description

A function to solve low rank plus sparse model estimation

Usage

fista.LpS(
  data,
  lambda,
  mu,
  alpha_L = 0.25,
  niter = 100,
  backtracking = TRUE,
  x.true = NULL
)

Arguments

data	A numeric dataset with size of n by p
lambda	A positive numeric value, indicating the tuning parameter for sparse component
mu	A positive numeric value, indicating the tuning parameter for low rank component
alpha_L	The constraint coefficient of low rank component, default is 0.25
niter	The maximum number of iterations required for FISTA
backtracking	A boolean argument, indicating that use backtracking in the FISTA
x.true	A p by p matrix, the true model parameter. Only available for simulation.

Value

A S3 object of class LSVAR, including

est_phi

estimated model parameter

sparse.comp

Estimated sparse component

lr.comp

Estimated low-rank component

obj.val

Values of objective function

rel.err

Relative errors compared with the true model parameters if available

Examples

n <- 300
p <- 20
try <- testVAR(n, p, struct = "LS", signal = 0.75, rank = 2,
               singular_vals = c(1, 0.8))
data <- as.matrix(try$series)
lambda <- 0.1; mu <- 1
fit <- fista.LpS(data, lambda = lambda, mu = mu, x.true = try$model_param)
summary(fit, threshold = 0.2)

LSVAR documentation built on May 26, 2021, 5:07 p.m.