Struct LogisticRegression 

#[non_exhaustive]
pub struct LogisticRegression { /* private fields */ }

Logistic regression for binary/multiclass classification.

Uses L-BFGS (default) or gradient descent with configurable learning rate, iterations, and L2 regularization.

Implementations

impl LogisticRegression

pub fn new() -> Self

Create a new logistic regression model.

pub fn learning_rate(self, lr: f64) -> Self

Set the learning rate (used by GradientDescent solver only).

pub fn max_iter(self, n: usize) -> Self

Set maximum iterations.

pub fn alpha(self, a: f64) -> Self

Set regularization strength (equivalent to 1/C in scikit-learn).

The meaning depends on the Penalty:

• L2 / L1 — multiplier on the penalty term
• ElasticNet — total regularization strength (split by l1_ratio)
• None — ignored

To match scikit-learn’s LogisticRegression(C=x), use alpha(1.0 / x). The default alpha = 1.0 corresponds to C = 1.0.

pub fn penalty(self, p: Penalty) -> Self

Set the regularization penalty.

Default is Penalty::L2. Use Penalty::L1 for sparse feature selection.

Errors

L1 and ElasticNet are not supported with the Lbfgs solver — calling fit() will return Err(InvalidParameter). Switch to GradientDescent.

pub fn tolerance(self, t: f64) -> Self

Set convergence tolerance.

pub fn tol(self, t: f64) -> Self

Alias for tolerance (sklearn convention).

pub fn class_weight(self, cw: ClassWeight) -> Self

Set class weighting strategy for imbalanced datasets.

pub fn solver(self, s: Solver) -> Self

Set the solver algorithm.

Defaults to Solver::Lbfgs which is ~10-20× faster than gradient descent.

pub fn fit(&mut self, data: &Dataset) -> Result<()>

Train the model using the configured solver.

Uses consistent softmax for both training and inference (not one-vs-rest sigmoid).

Errors

Returns InvalidParameter if Penalty::L1 or Penalty::ElasticNet is used with the Lbfgs solver (L-BFGS requires a differentiable objective).

pub fn predict(&self, features: &[Vec<f64>]) -> Result<Vec<f64>>

Predict class labels.

pub fn predict_proba(&self, features: &[Vec<f64>]) -> Result<Vec<Vec<f64>>>

Predict class probabilities.

pub fn fit_sparse(&mut self, features: &CscMatrix, target: &[f64]) -> Result<()>

Fit on sparse features using gradient descent.

Accepts CscMatrix (column-oriented) for efficient gradient computation. Only supports L2 penalty (or None). Uses gradient descent (not L-BFGS).

pub fn predict_sparse(&self, features: &CsrMatrix) -> Result<Vec<f64>>

Predict class labels from sparse features (CSR format).

pub fn predict_proba_sparse( &self, features: &CsrMatrix, ) -> Result<Vec<Vec<f64>>>

Predict class probabilities from sparse features (CSR format).

pub fn weights(&self) -> &[Vec<f64>]

Get learned weights (coefficients + bias) for each class.

Trait Implementations

impl CalibrableClassifier for LogisticRegression

fn fit(&mut self, data: &Dataset) -> Result<()>

Train on a dataset.

fn predict(&self, features: &[Vec<f64>]) -> Result<Vec<f64>>

Predict class labels.

fn predict_proba(&self, features: &[Vec<f64>]) -> Result<Vec<Vec<f64>>>

Predict class probabilities. Returns [n_samples][n_classes].

fn clone_box(&self) -> Box<dyn CalibrableClassifier>

Clone into a boxed trait object.

impl Clone for LogisticRegression

fn clone(&self) -> LogisticRegression

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Default for LogisticRegression

fn default() -> Self

Returns the “default value” for a type.

impl PartialFit for LogisticRegression

fn partial_fit(&mut self, data: &Dataset) -> Result<()>

Run one pass of gradient descent on the given batch.

On the first call, initializes weights from the data dimensions and class count. Subsequent calls preserve weights and continue updating.

fn is_initialized(&self) -> bool

Whether the model has been initialized (at least one partial_fit call).

impl PipelineModel for LogisticRegression

fn fit(&mut self, data: &Dataset) -> Result<()>

Train the model on a dataset.

fn predict(&self, features: &[Vec<f64>]) -> Result<Vec<f64>>

Predict on row-major feature matrix.

impl Tunable for LogisticRegression

fn set_param(&mut self, name: &str, _value: ParamValue) -> Result<()>

Apply a named hyperparameter.

fn clone_box(&self) -> Box<dyn Tunable>

Clone this model into a boxed trait object.

fn fit(&mut self, data: &Dataset) -> Result<()>

Train on a dataset.

fn predict(&self, features: &[Vec<f64>]) -> Result<Vec<f64>>

Predict on row-major features.