Creates explainer of PAN model

Source: R/explain_pan.R

Creates DALEX explainer for PAN model (or other neural network) All DALEX functions such as model_performance are possible to use on the returned explainer.

explain_pan(
  y,
  model,
  label,
  original_data,
  data,
  batch_size,
  dev,
  verbose = TRUE
)

Arguments

y

numerical target of classification task
model

net, nn_module, the model we want to explain
label

character providing the label (name) to first model
original_data

numerical list (table) of predictors
data

scaled matrix of numerical values representing predictors
batch_size

integer indicating a batch size used in dataloader.
dev

device used to calculations (cpu or gpu)
verbose

logical indicating if we want to print monitored outputs or not. Default: TRUE.

Value

DALEX model explainer

Examples


if (FALSE) {
dev <- "cpu"

adult <- fairmodels::adult

processed <- preprocess(
  adult,
  "salary",
  "sex",
  "Male",
  "Female",
  c("race"),
  sample = 0.8,
  train_size = 0.65,
  test_size = 0.35,
  validation_size = 0,
  seed = 7
)
# presaved torch model
model1 <- torch::torch_load(system.file("extdata","clf1",package="fairpan"))

dsl <- dataset_loader(processed$train_x, processed$train_y,
                      processed$test_x, processed$test_y,
                      batch_size=5, dev=dev)

explainer <- explain_pan(
  processed$test_y,
  model1,
  "classifier",
  processed$data_test,
  processed$data_scaled_test,
  batch_size = 5,
  dev = dev,
  verbose = FALSE
)
}