Struct IterativeImputer 

pub struct IterativeImputer<F> { /* private fields */ }

An unfitted iterative imputer.

Calling Fit::fit learns the per-feature BayesianRidge models and returns a FittedIterativeImputer that can impute missing values in new data by deterministically replaying the learned imputation sequence.

Parameters

• max_iter — maximum number of imputation rounds (default 10).
• tol — convergence tolerance on the inf-norm of the change scaled by max|X_observed| (default 1e-3).
• initial_strategy — strategy for the initial fill (default Mean).
• imputation_order — feature visit order (default Ascending).
• min_value / max_value — imputed-value clip bounds (default ±∞).

Examples

use ferrolearn_preprocess::iterative_imputer::{IterativeImputer, InitialStrategy};
use ferrolearn_core::traits::{Fit, Transform};
use ndarray::array;

let imputer = IterativeImputer::<f64>::new(10, 1e-3, InitialStrategy::Mean);
let x = array![[1.0, 2.0], [3.0, f64::NAN], [f64::NAN, 6.0]];
let fitted = imputer.fit(&x, &()).unwrap();
let out = fitted.transform(&x).unwrap();
assert!(!out[[1, 1]].is_nan());
assert!(!out[[2, 0]].is_nan());

Implementations

impl<F: Float + Send + Sync + 'static> IterativeImputer<F>

pub fn new(max_iter: usize, tol: F, initial_strategy: InitialStrategy) -> Self

Create a new IterativeImputer with the given core parameters and the sklearn defaults for imputation_order (Ascending) and the min_value/max_value clip (-inf/+inf).

pub fn with_imputation_order(self, order: ImputationOrder) -> Self

Set the feature imputation order (default Ascending), mirroring sklearn’s imputation_order (_iterative.py:316).

pub fn with_min_value(self, min_value: F) -> Self

Set the minimum imputed value (the clip lower bound), mirroring sklearn’s min_value (_iterative.py:318, default -inf).

pub fn with_max_value(self, max_value: F) -> Self

Set the maximum imputed value (the clip upper bound), mirroring sklearn’s max_value (_iterative.py:319, default +inf).

pub fn max_iter(&self) -> usize

Return the maximum number of iterations.

pub fn tol(&self) -> F

Return the convergence tolerance.

pub fn initial_strategy(&self) -> InitialStrategy

Return the initial imputation strategy.

pub fn imputation_order(&self) -> ImputationOrder

Return the feature imputation order.

pub fn min_value(&self) -> F

Return the minimum imputed value (clip lower bound).

pub fn max_value(&self) -> F

Return the maximum imputed value (clip upper bound).

Trait Implementations

impl<F: Clone> Clone for IterativeImputer<F>

fn clone(&self) -> IterativeImputer<F>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<F: Debug> Debug for IterativeImputer<F>

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

Formats the value using the given formatter.

impl<F: Float + Send + Sync + 'static> Default for IterativeImputer<F>

fn default() -> Self

Returns the “default value” for a type.

impl<F: ImputerFloat> Fit<ArrayBase<OwnedRepr<F>, Dim<[usize; 2]>>, ()> for IterativeImputer<F>

fn fit( &self, x: &Array2<F>, _y: &(), ) -> Result<FittedIterativeImputer<F>, FerroError>

Fit the iterative imputer by round-robin BayesianRidge regression, mirroring sklearn.impute.IterativeImputer.fit_transform (sklearn/impute/_iterative.py:693-831) in the sample_posterior=False path.

Features with missing values are visited in imputation_order (default Ascending); each is fit on the rows where it is observed and its missing rows are predicted and clipped to [min_value, max_value]. The round-robin repeats up to max_iter rounds, breaking early when max|Xt - Xt_prev| < tol * max|X_observed| (inf-norm convergence, :780,811,818).

Errors

• FerroError::InsufficientSamples if the input has zero rows.
• Propagates FerroError from the per-feature BayesianRidge fit.

max_iter == 0 is valid (matching sklearn _iterative.py:750-752): the iteration loop runs zero times and fit returns the initial fill with n_iter() == 0.

type Fitted = FittedIterativeImputer<F>

The fitted model type returned by fit.

type Error = FerroError

The error type returned by fit.

impl<F: ImputerFloat> FitTransform<ArrayBase<OwnedRepr<F>, Dim<[usize; 2]>>> for IterativeImputer<F>

fn fit_transform(&self, x: &Array2<F>) -> Result<Array2<F>, FerroError>

Fit the imputer on x and return the imputed output in one step.

Errors

Returns an error if fitting fails.

type FitError = FerroError

The error type for the combined fit-transform operation.

impl<F: Float + Send + Sync + 'static> Transform<ArrayBase<OwnedRepr<F>, Dim<[usize; 2]>>> for IterativeImputer<F>

Implement Transform on the unfitted imputer to satisfy the FitTransform: Transform supertrait bound.

fn transform(&self, _x: &Array2<F>) -> Result<Array2<F>, FerroError>

Always returns an error — the imputer must be fitted first.

type Output = ArrayBase<OwnedRepr<F>, Dim<[usize; 2]>>

The transformed output type.

type Error = FerroError

The error type returned by transform.