Struct ConditionalTabularMasker 

pub struct ConditionalTabularMasker { /* private fields */ }

Empirical conditional tabular masker using nearest background rows on the currently present features. Categorical columns use exact-match distance; numerical columns use variance-scaled squared distance.

Implementations

impl ConditionalTabularMasker

pub fn new( background: Background, categorical_features: &[usize], neighbors: usize, ) -> Result<Self>

pub fn background(&self) -> &Background

pub fn neighbors(&self) -> usize

Trait Implementations

impl Clone for ConditionalTabularMasker

fn clone(&self) -> ConditionalTabularMasker

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ConditionalTabularMasker

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Masker for ConditionalTabularMasker

fn n_features(&self) -> usize

Number of coalition features exposed to an explainer.

fn mask( &self, sample: ArrayView1<'_, f64>, present: &[bool], ) -> Result<Array2<f64>>

fn n_input_features(&self) -> usize

Number of columns expected in model input samples.

fn attribution_data(&self, samples: ArrayView2<'_, f64>) -> Result<Array2<f64>>

Values stored on the feature axis of the resulting explanation.

fn for_each_masked_batch( &self, sample: ArrayView1<'_, f64>, present: &[bool], visitor: &mut dyn FnMut(Array2<f64>) -> Result<()>, ) -> Result<()>

Visits model-input batches for one coalition.

fn streams_masked_batches(&self) -> bool

Whether Masker::for_each_masked_batch should be consumed directly instead of participating in ordinary multi-coalition batching.