Default Loss Metric for an Outcome Type

Returns the loss metric used for reporting cross-validated / cross-fitted empirical loss when the user does not supply one.

Usage

default_loss_metric(
  outcome_type = c("continuous", "binary", "density", "multiclass")
)

Arguments

outcome_type

One of 'continuous', 'binary', 'density', or 'multiclass'.

Value

A loss function.

Examples

# the loss used for cross-validated reporting when none is supplied;
# for continuous outcomes this is mean squared error:
continuous_loss <- default_loss_metric("continuous")
continuous_loss(c(1.5, 2.0), c(1, 2))  # (predicted, observed)
#> [1] 0.125

# for binary outcomes, negative log loss:
binary_loss <- default_loss_metric("binary")
binary_loss(c(0.9, 0.2, 0.8), c(1, 0, 1))
#> [1] 0.5516476