Struct FittedComplementNB 

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

Fitted Complement Naive Bayes classifier.

Implementations

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

pub fn partial_fit( &mut self, x: &Array2<F>, y: &Array1<usize>, ) -> Result<(), FerroError>

Incrementally update the model with new data.

Accumulates feature counts and class counts, then recomputes the complement weights.

Errors

• FerroError::ShapeMismatch if x and y have different row counts or the number of features does not match the fitted model.
• FerroError::InvalidParameter if any feature value is negative.

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

Predict class probabilities for the given feature matrix.

Returns shape (n_samples, n_classes) where each row sums to 1. Delegates to BaseNB::nb_predict_proba — with ComplementNB’s sklearn-parity sign, the joint log-likelihood is X @ weights.T directly, so exp(jll - logsumexp(jll)) is the softmax of the complement scores.

Errors

Returns FerroError::ShapeMismatch if the number of features does not match the fitted model.

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

Compute the joint log-likelihood scores using sklearn’s sign convention: argmax(jll) gives the predicted class.

Returns shape (n_samples, n_classes). With the sklearn-parity sign, X @ weights.T IS the joint log-likelihood. Matches sklearn ComplementNB._joint_log_likelihood. Delegates to BaseNB::nb_predict_joint_log_proba.

Errors

Returns FerroError::ShapeMismatch if the number of features does not match the fitted model.

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

Compute log of class probabilities (numerically stable).

Returns shape (n_samples, n_classes). Delegates to BaseNB::nb_predict_log_proba.

Errors

Returns FerroError::ShapeMismatch if the number of features does not match the fitted model.

pub fn score(&self, x: &Array2<F>, y: &Array1<usize>) -> Result<F, FerroError>

Mean accuracy on the given test data and labels.

Equivalent to sklearn’s ClassifierMixin.score.

Errors

Returns FerroError::ShapeMismatch if x.nrows() != y.len() or the feature count does not match the fitted model.

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

pub fn feature_log_prob(&self) -> &Array2<F>

Empirical complement weights (the negated smoothed complement-class log-probabilities), shape (n_classes, n_features).

Mirrors sklearn ComplementNB.feature_log_prob_ (_update_feature_log_prob, naive_bayes.py:1042).

pub fn feature_count(&self) -> &Array2<F>

Number of samples encountered for each (class, feature) during fitting, shape (n_classes, n_features).

Mirrors sklearn ComplementNB.feature_count_ (_count, naive_bayes.py:961).

pub fn class_count(&self) -> Array1<F>

Number of samples encountered for each class during fitting, shape (n_classes,).

Mirrors sklearn ComplementNB.class_count_ (_count, naive_bayes.py:951). class_counts is stored as integer counts; this casts each to F.

pub fn feature_all(&self) -> Array1<F>

Number of samples encountered for each feature during fitting (the per-feature total across all classes), shape (n_features,).

Derived (not stored) as feature_count_.sum(axis=0), mirroring sklearn ComplementNB.feature_all_ (_count, feature_all_ = feature_count_.sum(axis=0), naive_bayes.py:1029).

pub fn class_log_prior(&self) -> Array1<F>

Smoothed empirical log probability for each class, shape (n_classes,).

Derived (not stored) as log(class_count_) - log(class_count_.sum()), mirroring sklearn’s EMPIRICAL class_log_prior_ under the default fit_prior=True (_update_class_log_prior, naive_bayes.py:600). ComplementNB stores the empirical class-prior derivation; this returns the EMPIRICAL prior (matching sklearn’s class_log_prior_ value on any fit). Note: ComplementNB only consults class_log_prior_ in the single-class edge case (naive_bayes.py:1047-1048); it does not affect multi-class predictions.

Trait Implementations

impl<F: Float + Send + Sync + 'static> BaseNB<F> for FittedComplementNB<F>

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

Compute the joint log-likelihood scores for each class — sklearn ComplementNB._joint_log_likelihood.

Returns X @ feature_log_prob_.T (shape (n_samples, n_classes)). With ferrolearn’s sklearn-parity sign for feature_log_prob_, higher is better and argmax(scores, axis=1) predicts the class.

fn nb_classes(&self) -> &[usize]

The sorted class labels — the classes_ attribute.

fn nb_predict(&self, x: &Array2<F>) -> Result<Array1<usize>, FerroError>

Predict class labels: classes_[argmax(jll, axis=1)].

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

Return log-probability estimates: jll - logsumexp(jll, axis=1).

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

Return probability estimates: exp(predict_log_proba).

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

Return the unnormalized joint log-probability estimates.

impl<F: Clone> Clone for FittedComplementNB<F>

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

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<F: Debug> Debug for FittedComplementNB<F>

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

Formats the value using the given formatter.

impl<F: Float + Send + Sync + 'static> HasClasses for FittedComplementNB<F>

fn classes(&self) -> &[usize]

Returns the sorted list of unique class labels.

fn n_classes(&self) -> usize

Returns the number of distinct classes.

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

fn predict(&self, x: &Array2<F>) -> Result<Array1<usize>, FerroError>

Predict class labels for the given feature matrix.

With ComplementNB’s sklearn-parity sign, the highest joint log-likelihood wins. Delegates to BaseNB::nb_predict.

Errors

Returns FerroError::ShapeMismatch if the number of features does not match the fitted model.

type Output = ArrayBase<OwnedRepr<usize>, Dim<[usize; 1]>>

The prediction output type (e.g., ndarray::Array1<F> or ndarray::Array1<usize>).

type Error = FerroError

The error type returned by predict.