Struct BayesianLogisticRegression 

pub struct BayesianLogisticRegression { /* private fields */ }

Bayesian Logistic Regression with Laplace approximation.

Uses a Gaussian approximation to the posterior distribution around the MAP (Maximum A Posteriori) estimate.

Example

use aprender::bayesian::BayesianLogisticRegression;
use aprender::primitives::{Matrix, Vector};

let mut model = BayesianLogisticRegression::new(1.0); // precision = 1.0
model.fit(&x_train, &y_train).expect("fit should succeed with valid data");

let probas = model.predict_proba(&x_test).expect("prediction should succeed after fitting");
let (lower, upper) = model.predict_proba_interval(&x_test, 0.95).expect("interval prediction should succeed after fitting");

Implementations

impl BayesianLogisticRegression

pub fn new(prior_precision: f32) -> Self

Creates a new Bayesian Logistic Regression with specified prior precision.

Arguments

• prior_precision - Precision λ of the Gaussian prior (λ = 1/σ²)
  • Higher precision = stronger regularization
  • Typical values: 0.1 - 10.0

Example

use aprender::bayesian::BayesianLogisticRegression;

let model = BayesianLogisticRegression::new(1.0);

pub fn with_learning_rate(self, lr: f32) -> Self

Sets the learning rate for MAP estimation.

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

Sets the maximum number of iterations.

pub fn with_tolerance(self, tol: f32) -> Self

Sets the convergence tolerance.

pub fn coefficients_map(&self) -> Option<&[f32]>

Returns the MAP estimate of coefficients (available after fitting).

pub fn posterior_covariance(&self) -> Option<&[Vec<f32>]>

Returns the posterior covariance matrix (available after fitting).

pub fn fit(&mut self, x: &Matrix<f32>, y: &Vector<f32>) -> Result<()>

Fits the Bayesian Logistic Regression using Laplace approximation.

Finds the MAP estimate β_MAP and computes the Hessian for uncertainty.

Arguments

• x - Feature matrix (n × p)
• y - Binary labels (n), must be 0.0 or 1.0

Returns

Ok(()) on success, error if dimensions mismatch or convergence fails

pub fn predict_proba(&self, x_test: &Matrix<f32>) -> Result<Vector<f32>>

Predicts class probabilities for test data.

Returns the probability of class 1 for each sample.

Arguments

• x_test - Test feature matrix (n_test × p)

Returns

Vector of probabilities P(y=1 | x)

pub fn predict(&self, x_test: &Matrix<f32>) -> Result<Vector<f32>>

Predicts class labels (0 or 1) for test data.

Uses threshold 0.5: predict 1 if P(y=1|x) >= 0.5, else 0.

pub fn predict_proba_interval( &self, x_test: &Matrix<f32>, level: f32, ) -> Result<(Vector<f32>, Vector<f32>)>

Predicts class probabilities with credible intervals.

Uses the Laplace approximation to compute Bayesian credible intervals for predicted probabilities.

Arguments

• x_test - Test feature matrix (n_test × p)
• level - Confidence level (e.g., 0.95 for 95% credible interval)

Returns

Tuple of (lower_bounds, upper_bounds) for P(y=1|x)

Example

let (lower, upper) = model.predict_proba_interval(&x_test, 0.95)?;

Trait Implementations

impl Clone for BayesianLogisticRegression

fn clone(&self) -> BayesianLogisticRegression

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for BayesianLogisticRegression

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

Formats the value using the given formatter.