WInfL1: Infinity-Wasserstein Linear Projections With an L1 Penalty

Description

Infinity-Wasserstein Linear Projections With an L1 Penalty

Usage

WInfL1(
  X,
  Y,
  theta = NULL,
  penalty = c("none", "lasso", "mcp", "scad"),
  lambda = numeric(0),
  lambda.min.ratio = 1e-04,
  gamma = 1.5,
  nlambda = 10,
  solver = c("cone", "mosek", "gurobi"),
  options = list(solver_opts = NULL, init = NULL, tol = 1e-07, iter = 100),
  model.size = NULL,
  display.progress = FALSE,
  ...
)

Value

A WpProj object

Arguments

X: An n x p matrix of covariates
Y: An n x s matrix of predictions
theta: optional parameter matrix for selection methods. Should be p x s.
penalty: Form of penalty. One of "none", "lasso", "mcp","scad"
lambda: Penalty parameter for lasso regression.
lambda.min.ratio: Minimum lambda ratio for self selected lambda.
gamma: tuning parameters for SCAD and MCP.
nlambda: Number of lambda values.
solver: Which solver to use. One of "cone","mosek", or "gurobi". Note "mosek" and "gurobi" are commercial installers.
options: A list containing slots solver_opts, options for each solver, init, initial conditions fed into each solver, tol, tolerance for convergence, and iter, the maximum number of iterations
model.size: The maximum number of paramters to consider. Should be an integer greater than 1 and less than or equal to the number of covariates
display.progress: Whether to display progress. TRUE or FALSE
...: Additional arguments passed to the solver as needed