tests/testthat/test-LinearDiscriminantClassifier.R

In RSSL: Implementations of Semi-Supervised Learning Approaches for Classification

context("Linear Discriminant Classifier")

# Simple dataset used in the tests
data(testdata)
modelform <- testdata$modelform
classname<-all.vars(modelform)[1] 
D <- testdata$D
D_test <- testdata$D_test
X <- testdata$X
X_u <- testdata$X_u
y <- testdata$y
X_test <- testdata$X_test
y_test <- testdata$y_test

test_that("Different input schemes", {
  g_matrix <- LinearDiscriminantClassifier(X,y)
  g_model <- LinearDiscriminantClassifier(modelform, D)
  
  expect_equal(predict(g_matrix, X_test),predict(g_model, D_test))
  expect_equal(loss(g_matrix, X_test, y_test),loss(g_model, D_test))
})

test_that("Scaling input has some expected properties", {
  g_unscaled <- LinearDiscriminantClassifier(X, y, x_center=TRUE, scale=FALSE)
  g_unscaled2 <- LinearDiscriminantClassifier(X, y, x_center=FALSE, scale=FALSE)
  g_scaled <- LinearDiscriminantClassifier(X, y, x_center=TRUE, scale=TRUE)
  
  expect_equal(predict(g_scaled, X_test),predict(g_unscaled, X_test))
  expect_equal(predict(g_unscaled2, X_test),predict(g_unscaled, X_test))
  
  expect_equal(predict(g_scaled, X_test)=="-1",
               posterior(g_scaled, X_test)[,1]>=0.5)
  expect_equal(predict(g_unscaled, X_test)=="-1",
               posterior(g_unscaled, X_test)[,1]>=0.5)
  
  expect_equal(loss(g_scaled, X_test, y_test)-loss(g_unscaled, X_test, y_test),rep(-0.9848311,nrow(X_test)),tolerance=10e-7)
  expect_equal(loss(g_unscaled, X_test, y_test),loss(g_unscaled2, X_test, y_test))
  
  expect_equal(posterior(g_scaled, X_test)[,1],posterior(g_unscaled, X_test)[,1], tolerance=10e-5)
  expect_equal(posterior(g_unscaled2, X_test),posterior(g_unscaled, X_test))
  
  expect_equal(line_coefficients(g_scaled), line_coefficients(g_unscaled))
  expect_equal(line_coefficients(g_unscaled), line_coefficients(g_unscaled2))
})