W2IP: 2-Wasserstein distance selection by Integer Programming

In WpProj: Linear p-Wasserstein Projections

2-Wasserstein distance selection by Integer Programming

Description

2-Wasserstein distance selection by Integer Programming

Usage

W2IP(
  X,
  Y = NULL,
  theta,
  transport.method = transport_options(),
  model.size = NULL,
  nvars = NULL,
  maxit = 100L,
  infimum.maxit = 100L,
  tol = 1e-07,
  solver = c("cone", "lp", "mosek", "cplex", "gurobi"),
  display.progress = FALSE,
  parallel = NULL,
  ...
)

Arguments

X	Covariates
Y	Predictions from arbitrary model
theta	Parameters of original linear model. Required
transport.method	Method for Wasserstein distance calculation. Should be one of the outputs of transport_options().
model.size	Maximum number of coefficients in interpretable model
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 solver iterations
infimum.maxit	Maximum iterations to alternate binary program and Wasserstein distance calculation
tol	Tolerance for convergence of coefficients
solver	The solver to use. Must be one of "cone","lp", "cplex", "gurobi","mosek".
display.progress	Should progress be printed?
parallel	foreach back end. See foreach::foreach() for more details.
...	Extra args to Wasserstein distance methods

Details

For argument solution.method, options "cone" and "lp" use the free solvers "ECOS" and "lpSolver", respectively. "cplex", "gurobi" and "mosek" require installing the corresponding commercial solvers.

WpProj documentation built on May 29, 2024, 7:55 a.m.