tests/testthat/test_optimizer.R

In schalkdaniel/compboost: Efficient Component-Wise Boosting Implementation

context("Optimizer of 'compboost'")

test_that("Coordinate Descent with line search works", {
  n_train = 1000L

  expect_silent({ used_optimizer = OptimizerCoordinateDescent$new() })

  cboost = Compboost$new(data = mtcars, target = "mpg", optimizer = used_optimizer, loss = LossQuadratic$new(), learning_rate = 0.05)

  cboost$addBaselearner("wt", "linear", BaselearnerPolynomial)
  cboost$addBaselearner("disp", "linear", BaselearnerPolynomial)
  cboost$addBaselearner("hp", "linear", BaselearnerPolynomial)

  nuisance = capture.output(suppressWarnings({
    cboost$train(n_train)
  }))

  used_optimizer_ls = OptimizerCoordinateDescentLineSearch$new()

  nuisance = capture.output(suppressWarnings({
    cboost1 = Compboost$new(data = mtcars, target = "mpg", optimizer = used_optimizer_ls, loss = LossQuadratic$new(), learning_rate = 0.05)
  }))

  cboost1$addBaselearner("wt", "linear", BaselearnerPolynomial)
  cboost1$addBaselearner("disp", "linear", BaselearnerPolynomial)
  cboost1$addBaselearner("hp", "linear", BaselearnerPolynomial)

  nuisance = capture.output(suppressWarnings({
    cboost1$train(n_train)
  }))

  expect_equal(cboost$predict(), cboost1$predict())
  expect_equal(cboost$getInbagRisk(), cboost1$getInbagRisk())
  expect_true(all(abs(used_optimizer_ls$getStepSize() - 1) < 1e10))
  expect_true(var(used_optimizer_ls$getStepSize()) < 1e-10)
})

test_that("AGBM optimizer works", {
  n_train = 1000L

  expect_silent({ used_optimizer = OptimizerAGBM$new(0.2) })

  expect_silent({
    cboost = Compboost$new(data = mtcars, target = "mpg", optimizer = used_optimizer, loss = LossQuadratic$new(),
      learning_rate = 0.05)

    cboost$addBaselearner("wt", "linear", BaselearnerPolynomial)
    cboost$addBaselearner("disp", "linear", BaselearnerPolynomial)
    cboost$addBaselearner("hp", "linear", BaselearnerPolynomial)
  })
  nuisance = capture.output(suppressWarnings({
    cboost$train(n_train)
  }))

  expect_equal(length(used_optimizer$getSelectedMomentumBaselearner()), n_train)
  expect_equal(cboost$predict(mtcars), cboost$predict())
  expect_equal(used_optimizer$getStepSize(), 1)
  expect_equal(cboost$response$getPrediction(), cboost$predict(mtcars))

  expect_silent({ used_optimizer2 = OptimizerAGBM$new(0.2) })
  expect_silent({
    cboost2 = Compboost$new(data = mtcars, target = "mpg", optimizer = used_optimizer2, loss = LossQuadratic$new(),
      learning_rate = 0.05)

    cboost2$addBaselearner("wt", "linear", BaselearnerPolynomial)
    cboost2$addBaselearner("disp", "linear", BaselearnerPolynomial)
    cboost2$addBaselearner("hp", "linear", BaselearnerPolynomial)
  })
  nuisance = capture.output(suppressWarnings({
    cboost2$train(1500)
  }))
  expect_output(cboost$train(1500))
  expect_equal(cboost$predict(), cboost2$predict())
})