Functional Data Depth Features

Computes the data depth of functional features via tf::tf_depth(). Data depth measures how central each curve is relative to the others: values close to 1 indicate central curves and values close to 0 indicate extreme curves.

The depth is computed in-sample, i.e. each curve is scored relative to the other curves in the same data. The same operation is applied during training and prediction.

Irregular curves are interpolated to a common grid before the depth is computed. Curves that remain incomplete after interpolation (e.g. because they only cover part of the grid) are assigned a depth of NA.

Details

The "MHI" method ranks functions from lowest (0) to highest (1) instead of from most extreme to most central. The "RPD" method relies on random projections, so set a seed (e.g. via set.seed()) before training for reproducible results.

Parameters

The parameters are the parameters inherited from PipeOpTaskPreprocSimple, as well as the following parameters:

• method :: character(1)
  The depth method to use. One of "MBD" (modified band depth, the default), "MHI" (modified hypograph index), "FM" (Fraiman-Muniz), "FSD" (functional spatial depth), or "RPD" (regularized projection depth). Initial value is "MBD".

• na.rm :: logical(1)
  Whether to remove missing observations before computing the depth. Initial value is TRUE.

• u :: numeric(1)
  Quantile level for the regularization. Only used when method is "RPD". Default is 0.01.

• n_projections :: integer(1)
  Number of projection directions. Only used when method is "RPD". Default is 5000.

• n_projections_beta :: integer(1)
  Number of directions for estimating the regularization parameter. Only used when method is "RPD". Default is 500.

Naming

The new names generally append a _depth to the corresponding column name. If a column was called "x", the corresponding new column will be called "x_depth".

Super classes

mlr3pipelines::PipeOp -> mlr3pipelines::PipeOpTaskPreproc -> mlr3pipelines::PipeOpTaskPreprocSimple -> PipeOpFDADepth

Methods

Public methods

• PipeOpFDADepth$new()

• PipeOpFDADepth$clone()

Inherited methods

• mlr3pipelines::PipeOp$help()
• mlr3pipelines::PipeOp$predict()
• mlr3pipelines::PipeOp$print()
• mlr3pipelines::PipeOp$train()

---

PipeOpFDADepth$new()

Initializes a new instance of this Class.

Usage

PipeOpFDADepth$new(id = "fda.depth", param_vals = list())

Arguments

id

(character(1))
Identifier of resulting object, default "fda.depth".

param_vals

(named list())
List of hyperparameter settings, overwriting the hyperparameter settings that would otherwise be set during construction. Default list().

---

PipeOpFDADepth$clone()

The objects of this class are cloneable with this method.

Usage

PipeOpFDADepth$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

task = tsk("fuel")
po_depth = po("fda.depth", method = "MBD")
task_depth = po_depth$train(list(task))[[1L]]
task_depth$data(cols = c("NIR_depth", "UVVIS_depth"))
#>       NIR_depth UVVIS_depth
#>           <num>       <num>
#>   1: 0.06086641  0.18607019
#>   2: 0.12161695  0.31235155
#>   3: 1.00176051  0.88750310
#>   4: 0.96914866  0.68872490
#>   5: 0.75128286  0.38775354
#>  ---                       
#> 125: 0.66711535  0.57225367
#> 126: 0.45788517  0.17058908
#> 127: 0.34026479  0.09003249
#> 128: 0.56095330  0.56875747
#> 129: 0.65649751  0.54263748