WpProj (version 0.2.1)

WPL1: p-Wasserstein Linear Projections With an \(L_1\) Penalty

Description

p-Wasserstein Linear Projections With an \(L_1\) Penalty

Usage

WPL1(
  X,
  Y = NULL,
  theta = NULL,
  power = 2,
  penalty = c("lasso", "ols", "mcp", "elastic.net", "selection.lasso", "scad", "mcp.net",
    "scad.net", "grp.lasso", "grp.lasso.net", "grp.mcp", "grp.scad", "grp.mcp.net",
    "grp.scad.net", "sparse.grp.lasso"),
  model.size = NULL,
  lambda = numeric(0),
  nlambda = 100L,
  lambda.min.ratio = 1e-04,
  gamma = 1,
  maxit = 500L,
  tol = 1e-07,
  ...
)

Value

object of class WpProj

Arguments

X

matrix of covariates

Y

matrix of predictions

theta

optional parameter matrix for selection methods.

power

power of the Wasserstein distance

penalty

Form of penalty. One of "lasso", "ols", "mcp", "elastic.net","selection.lasso", "scad", "mcp.net", "scad.net", "grp.lasso", "grp.lasso.net", "grp.mcp","grp.scad", "grp.mcp.net", "grp.scad.net", "sparse.grp.lasso"

model.size

How many coefficients should final model have

lambda

penalty parameter

nlambda

number of lambdas to explore

lambda.min.ratio

minimum ratio of max to min lambda

gamma

Tuning parameter for SCAD and MCP methods

maxit

maximum iterations for optimization

tol

tolerance for convergence

...

arguments passed to other methods such as Wasserstein distance

See Also

W1L1(), W2L1(), WInfL1