WPSA: p-Wasserstein distance projections using simulated annealing

Description

p-Wasserstein distance projections using simulated annealing

Usage

WPSA(
  X,
  Y = NULL,
  theta,
  power = 2,
  force = NULL,
  model.size = NULL,
  nvars = NULL,
  maxit = 1,
  temps = 1000,
  max.time = 3600,
  const = NULL,
  proposal = proposal.fun,
  options = list(method = c("selection.variable", "scale", "projection"),
    transport.method = transport_options(), energy.distribution = "boltzman",
    cooling.schedule = "Geman-Geman", proposal.method = "covariance", epsilon = 0.05,
    OTmaxit = 0),
  display.progress = FALSE,
  parallel = NULL,
  calc.theta = TRUE,
  xtx = NULL,
  xty = NULL,
  ...
)

Value

An object of class WpProj

Arguments

X: Covariate vector
Y: Predictions
theta: Optional matrix of parameters for generating predictions
power: Power of the Wasserstein distance
force: Any covariates to force into the model?
model.size: Maximum number of coefficients
nvars: The number of variables to explore. Should be an integer vector of model sizes. Default is NULL which will explore all models from 1 to model.size.
maxit: Maximum number of iterations
temps: Number of temperatures
max.time: Maximum time in seconds to run
const: Maximum value for simulated annealing distance
proposal: Proposal function. There is a default method but can provide your own with parameters xty, cur, idx, force, d, method
options: Options for simulated annealing
display.progress: Whether to display solver progress. TRUE or FALSE. Default is FALSE.
parallel: A foreach::foreach() backend
calc.theta: Should the model save the linear coefficients? TRUE or FALSE. Default is TRUE
xtx: precomputed crossproduct crossprod(X,X)
xty: precomputed crossprod(X, Y)