tests/testthat/test-admm.R

In straw-boy/solvers: Various solvers for SLOPE

test_that("ADMM: gaussian, n>p case", {
  library(SLOPE)
  set.seed(235)

  n = 100
  p = 10

  d <- randomProblem(n, p, response = "gaussian", density = 0.5)
  
  admm_solvers <- ADMM(d$x, d$y, family="gaussian",alpha=c(1.0,0.005),opt_algo="nr")
  fista_solvers <- FISTA(d$x, d$y, family="gaussian",alpha=c(1.0,0.005))
  expect_equivalent(coef(admm_solvers), coef(fista_solvers), tol = 1e-2)

})

test_that("ADMM: gaussian, n<p case", {

  library(SLOPE)
  set.seed(24)

  n = 10
  p = 20

  d <- randomProblem(n, p, response="gaussian", density = 0.5)
  
  admm_solvers <- ADMM(d$x, d$y, family="gaussian",alpha=c(1.0,0.005),opt_algo="nr")
  fista_solvers <- FISTA(d$x, d$y, family="gaussian",alpha=c(1.0,0.005))
  
  expect_equivalent(coef(admm_solvers), coef(fista_solvers), tol = 1e-2)

})

test_that("ADMM: binomial, n>p case", {
  
  library(SLOPE)
  set.seed(186)

  n = 100
  p = 10

  d <- randomProblem(n, p, response="binomial", density = 0.5)
  
  admm_solvers <- ADMM(d$x, d$y, family="binomial",alpha=c(1.0,0.005),opt_algo="nr")
  fista_solvers <- FISTA(d$x, d$y, family="binomial",alpha=c(1.0,0.005))
  expect_equivalent(coef(admm_solvers), coef(fista_solvers), tol = 1e-2)

})


test_that("ADMM: binomial, n<p case", {

  library(SLOPE)
  set.seed(1)

  n = 10
  p = 20

  d <- randomProblem(n, p, response="binomial", density = 0.5)
  
  admm_solvers <- ADMM(d$x, d$y, family="binomial",alpha=c(1.0,0.005),opt_algo="nr")
  fista_solvers <- FISTA(d$x, d$y, family="binomial",alpha=c(1.0,0.005),tol_infeas=0,tol_rel_gap=0,max_passes=400)
  expect_equivalent(coef(admm_solvers), coef(fista_solvers), tol = 1e-2)

})

test_that("ADMM: poisson, n>p case", {

  library(SLOPE)
  set.seed(271)

  n = 100
  p = 10

  d <- randomProblem(n,p,response="poisson")
  
  admm_solvers <- ADMM(d$x, d$y, family="poisson",alpha=c(1.0,0.005),opt_algo="nr")
  fista_solvers <- FISTA(d$x, d$y, family="poisson",alpha=c(1.0,0.005))
  expect_equivalent(coef(admm_solvers), coef(fista_solvers), tol = 1e-2)

})

test_that("ADMM: poisson, n<p case", {

  library(SLOPE)
  set.seed(7)

  n = 10
  p = 20

  d <- randomProblem(n, p, response="poisson", density = 0.5)
  
  admm_solvers <- ADMM(d$x, d$y, family="poisson",alpha=c(1.0,0.005),opt_algo="nr")
  fista_solvers <- FISTA(d$x, d$y, family="poisson",alpha=c(1.0,0.005))
  expect_equivalent(coef(admm_solvers), coef(fista_solvers), tol = 1e-2)

})

test_that("ADMM: multinomial, n>p case", {

  library(SLOPE)
  set.seed(12)

  n = 100
  p = 10

  d <- randomProblem(n, p, response="multinomial", density = 0.5)
  
  admm_solvers <- ADMM(d$x, d$y, family="multinomial",alpha=c(1.0,0.005),opt_algo="nr")
  fista_solvers <- FISTA(d$x, d$y, family="multinomial",alpha=c(1.0,0.005))
  expect_equivalent(coef(admm_solvers), coef(fista_solvers), tol = 1e-2)

})

test_that("ADMM: multinomial, n<p case", {

  library(SLOPE)
  set.seed(1)

  n = 10
  p = 20

  d <- randomProblem(n, p, response="multinomial", density = 0.5)
  
  admm_solvers <- ADMM(d$x, d$y, family="multinomial",alpha=c(1.0,0.005),opt_algo="nr")
  fista_solvers <- FISTA(d$x, d$y, family="multinomial",alpha=c(1.0,0.005))
  expect_equivalent(coef(admm_solvers), coef(fista_solvers), tol = 1e-2)
  
})