tests/testthat/test-interpreter.R

In forestry-labs/interpretability_sandbox: Model Distillation and Interpretability Methods for Machine Learning Models

test_that("Tests that the interpreter wrapper is working", {

  library(Rforestry)
  set.seed(491)

  data <-iris

  test_ind <- sample(1:nrow(data), nrow(data)%/%5)
  train_reg <- data[-test_ind,]
  test_reg <- data[test_ind,]

  # Train a random forest on the data set
  forest <- forestry(x=train_reg[,-c(1)],
                     y=train_reg[,1])

  # Create a predictor wrapper for the forest
  forest_predictor <- Predictor$new(model = forest,
                                    data=train_reg,
                                    y="Sepal.Length",
                                    task = "regression")

  # Initialize an interpreter
  forest_interpret <- Interpreter$new(predictor = forest_predictor)

  context("Check the initialization of the interpreter object")

  expect_equal(length(forest_interpret$data.points[1:10]),
               10)

  expect_equal(forest_interpret$grid.size,
               50)

  expect_equal(forest_interpret$features,
               c("Sepal.Width",  "Petal.Length",
                 "Petal.Width",  "Species"     ))

  expect_equal(nrow(forest_interpret$features.2d),
               choose(length(forest_interpret$features), 2))

  expect_equal(all.equal(names(forest_interpret$center.at),
                         forest_interpret$features),
               TRUE)

  expect_equal(all.equal(forest_interpret$features,
                         names(forest_interpret$feat.class)),
               TRUE)

  expect_equal(all.equal(forest_interpret$features, names(forest_interpret$pdp.1d)),
               TRUE)

  expect_equal(all.equal(names(which(forest_interpret$feat.class != "factor")),
                         names(forest_interpret$ale.grid)),
               TRUE)

  expect_equal(all(is.na(forest_interpret$saved$ICE)), TRUE)

  expect_equal(all(is.na(forest_interpret$saved$PDP.1D)), TRUE)

  expect_equal(all(is.na(forest_interpret$saved$PDP.2D)), TRUE)


  # Get predictions from a pdp function
  context("Try pdp predictions")

  # Get the names of the pdp functions
  func_names <- names(forest_interpret$functions.1d)
  expect_equal(func_names, NULL)

  pred_pdp <- forest_interpret$pdp.1d$`Petal.Length`(c(10,11,12,14,15))
  pred_pdp_alt <- forest_interpret$pdp.1d[[2]](c(10,11,12,14,15))

  pred_pdp_2 <- forest_interpret$pdp.2d$Species$Petal.Length(data.frame(col1 = data$Species,
                                                                        col2 = data$Petal.Length))

  pred_pdp_2_alt <- forest_interpret$pdp.2d[[4]][[2]](data.frame(col1 = data$Species,
                                                               col2 = data$Petal.Length))

  expect_equal(all.equal(pred_pdp,pred_pdp_alt, tolerance = 1e-8), TRUE)
  expect_equal(all.equal(pred_pdp_2,pred_pdp_2_alt, tolerance = 1e-8), TRUE)

  expect_equal(length(pred_pdp), 5)
  expect_equal(length(pred_pdp_2), 150)

  rm(list=ls())
})