Loss Function

Represents a loss function for regression or classification tasks.

Public fields

id: (`character(1)`)
Identifier for the loss function.
label: (`character(1)`)
Label for the loss function.
task_type: (`character(1)`)
Task type: `"regr"` or `"classif"`.
input_default: (`character(1)`)
The natural input scale for the loss: `"score"` or `"probability"`.
input_supported: (`character()`)
All input scales this loss can be expressed on (subset of `c("score", "probability")`).
fun: (`function` or `list`)
The loss function itself. Can be a single function (for single input type) or a named list of functions for different input types.

Creates a new instance of this [R6][R6::R6Class] class.

LossFunction$new(
  id,
  label,
  task_type,
  fun,
  input_default = "score",
  input_supported = c("score")
)

id: (`character(1)`)
Identifier for the loss function.
label: (`character(1)`)
Label for the loss function.
task_type: (`character(1)`)
Task type: `"regr"` or `"classif"`.
fun: (`function` or `list`)
The loss function. Can be a single function or a named list of functions for different input types (e.g., list(score = function(r) ..., probability = function(pi) ...)).
input_default: (`character(1)`)
Default input scale (`"score"` or `"probability"`).
input_supported: (`character()`)
Character vector of supported input scales. Must contain `input_default`.

Get the loss function for a specific input type.

LossFunction$get_fun(input_type = self$input_default)

input_type: (`character(1)`)
The input type: `"score"` or `"probability"`.

A function for the specified input type.

The objects of this class are cloneable with this method.

LossFunction$clone(deep = FALSE)