Feature Sampler Class — FeatureSampler • xplainfi

Base class for implementing different sampling strategies for feature importance methods like PFI and CFI

Public fields

task: (mlr3::Task) Original task.
label: (character(1)) Name of the sampler.
feature_types: (character()) Feature types supported by the sampler. Will be checked against the provided mlr3::Task to ensure compatibility.
param_set: (paradox::ParamSet) Parameter set for the sampler.

Creates a new instance of the FeatureSampler class

FeatureSampler$new(task)

Sample values for feature(s) from stored task

FeatureSampler$sample(feature, row_ids = NULL, samples_per_row = 1L)

feature: (character) Feature name(s) to sample (can be single or multiple). Must match those in the stored Task.
row_ids: (integer(): NULL) Row IDs of the stored Task to use as basis for sampling.
samples_per_row: (integer(1): 1L) Number of independent samples to draw per input row. When > 1, the returned data.table has samples_per_row * length(row_ids) rows in draw-major order: rows 1..n are draw 1 across all input rows in order, rows n+1..2n are draw 2, and so on.

Modified copy of the input features with the feature(s) sampled. A data.table with samples_per_row * length(row_ids) rows in draw-major order.

Sample values for feature(s) using external data

FeatureSampler$sample_newdata(feature, newdata, samples_per_row = 1L)

feature: (character) Feature name(s) to sample (can be single or multiple)
newdata: (data.table) External data to use for sampling.
samples_per_row: (integer(1): 1L) See $sample(). Output is samples_per_row * nrow(newdata) rows in draw-major order.

Print sampler

FeatureSampler$print(...)

The objects of this class are cloneable with this method.

FeatureSampler$clone(deep = FALSE)