WpProj (version 0.2.3)

WPL0: p-Wasserstein projections with an L0 penalty

Description

p-Wasserstein projections with an L0 penalty

Usage

WPL0(
  X,
  Y = NULL,
  theta,
  power = 2,
  method = c("selection.variable", "projection"),
  transport.method = transport_options(),
  epsilon = 0.05,
  OTmaxit = 0,
  parallel = NULL,
  ...
)

Value

WpProj object

Arguments

X

matrix of covariates

Y

matrix of predictions

theta

optional matrix of coefficients from original model, if relevant

power

power of the Wasserstein distance

method

One of "selection.variable" or "projection". Methods decide whether covariate matrix in theta is preserved ("selection.variable") or if new projections are generated ("projection")

transport.method

Method for Wasserstein distance calculation. Should be one of the outputs of transport_options().

epsilon

hyperparameter for sinkhorn iterations

OTmaxit

max iteration for sinkhorn iterations

parallel

foreach backend