Struct VariationalSparseGaussianProcessRegressor 

pub struct VariationalSparseGaussianProcessRegressor<S = Untrained> { /* private fields */ }

let X = array![[0.0], [1.0], [2.0], [3.0], [4.0], [5.0], [6.0], [7.0]]; let y = array![0.0, 1.0, 4.0, 9.0, 16.0, 25.0, 36.0, 49.0];

let kernel = RBF::new(2.0); let vsgpr = VariationalSparseGaussianProcessRegressor::new() .kernel(Box::new(kernel)) .n_inducing(3); let fitted = vsgpr.fit(&X.view(), &y.view()).unwrap(); let predictions = fitted.predict(&X.view()).unwrap();

Implementations

impl VariationalSparseGaussianProcessRegressor<Untrained>

pub fn new() -> Self

Create a new VariationalSparseGaussianProcessRegressor instance

pub fn kernel(self, kernel: Box<dyn Kernel>) -> Self

Set the kernel function

pub fn n_inducing(self, n_inducing: usize) -> Self

Set the number of inducing points

pub fn inducing_init(self, inducing_init: InducingPointInit) -> Self

Set the inducing point initialization method

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

Set the learning rate for optimization

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

Set the maximum number of iterations

pub fn batch_size(self, batch_size: Option<usize>) -> Self

Set the batch size for mini-batch optimization

pub fn inducing_batch_size(self, inducing_batch_size: Option<usize>) -> Self

Set the inducing batch size for doubly stochastic optimization

pub fn sigma_n(self, sigma_n: f64) -> Self

Set the noise standard deviation

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

Set convergence tolerance

pub fn verbose(self, verbose: bool) -> Self

Set verbosity

pub fn random_state(self, random_state: Option<u64>) -> Self

Set the random state

pub fn optimizer(self, optimizer: VariationalOptimizer) -> Self

Set the optimization method

pub fn natural_gradient_damping(self, damping: f64) -> Self

Set the damping factor for natural gradients (only used when optimizer is NaturalGradients)

impl VariationalSparseGaussianProcessRegressor<VsgprTrained>

pub fn predict_with_std( &self, X: &ArrayView2<'_, f64>, ) -> SklResult<(Array1<f64>, Array1<f64>)>

Predict with uncertainty estimates

pub fn elbo(&self) -> f64

Get the evidence lower bound (ELBO)

pub fn elbo_history(&self) -> &[f64]

Get the ELBO history during training

pub fn inducing_points(&self) -> &Array2<f64>

Get the inducing points

pub fn variational_mean(&self) -> &Array1<f64>

Get the variational mean

pub fn variational_covariance(&self) -> &Array2<f64>

Get the variational covariance

pub fn update( self, X_new: &ArrayView2<'_, f64>, y_new: &ArrayView1<'_, f64>, learning_rate: Option<f64>, n_iterations: Option<usize>, ) -> SklResult<Self>

Online update with new data for streaming/scalable learning

This method allows incrementally updating the variational parameters with new data points without retraining from scratch, making it suitable for streaming scenarios and very large datasets.

Arguments

• X_new - New input data points
• y_new - New target values
• learning_rate - Learning rate for the update (if None, uses model’s learning rate)
• n_iterations - Number of update iterations (default: 10)

Returns

Updated model with modified variational parameters

pub fn recursive_update( self, X_new: &ArrayView2<'_, f64>, y_new: &ArrayView1<'_, f64>, forgetting_factor: Option<f64>, ) -> SklResult<Self>

Recursive Bayesian update for online GP learning

This method implements proper recursive Bayesian updates for Gaussian processes, maintaining the posterior mean and covariance through sequential updates without requiring full recomputation of the ELBO.

Arguments

• X_new - New input data (n_new x n_features)
• y_new - New target values (n_new,)
• forgetting_factor - Exponential forgetting factor (0 < λ ≤ 1)

Returns

Updated model with recursively updated posterior

pub fn sliding_window_update( self, X_new: &ArrayView2<'_, f64>, y_new: &ArrayView1<'_, f64>, window_size: usize, decay_rate: f64, ) -> SklResult<Self>

Sliding window update for streaming data

Maintains a sliding window of recent data points and uses exponential forgetting to down-weight older observations.

Arguments

• X_new - New input data
• y_new - New target values
• window_size - Maximum number of recent observations to maintain
• decay_rate - Exponential decay rate for older observations

pub fn adaptive_sparse_update( self, X_new: &ArrayView2<'_, f64>, y_new: &ArrayView1<'_, f64>, max_inducing: usize, quality_threshold: f64, removal_threshold: f64, ) -> SklResult<Self>

Adaptive sparse GP with dynamic inducing point management

This method adaptively adds/removes inducing points based on the approximation quality and computational budget, maintaining good approximation while controlling computational cost.

Arguments

• X_new - New input data
• y_new - New target values
• max_inducing - Maximum number of inducing points allowed
• quality_threshold - Minimum approximation quality threshold (0.0-1.0)
• removal_threshold - Threshold for removing redundant inducing points

Returns

Updated model with adaptively adjusted inducing points

Trait Implementations

impl<S: Clone> Clone for VariationalSparseGaussianProcessRegressor<S>

fn clone(&self) -> VariationalSparseGaussianProcessRegressor<S>

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<S: Debug> Debug for VariationalSparseGaussianProcessRegressor<S>

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

Formats the value using the given formatter.

impl Default for VariationalSparseGaussianProcessRegressor<Untrained>

fn default() -> Self

Returns the “default value” for a type.

impl Estimator for VariationalSparseGaussianProcessRegressor<Untrained>

type Config = GpcConfig

Configuration type for the estimator

type Error = SklearsError

Error type for the estimator

type Float = f64

The numeric type used by this estimator

fn config(&self) -> &Self::Config

Get estimator configuration

fn validate_config(&self) -> Result<(), SklearsError>

Validate estimator configuration with detailed error context

fn check_compatibility( &self, n_samples: usize, n_features: usize, ) -> Result<(), SklearsError>

Check if estimator is compatible with given data dimensions

fn metadata(&self) -> EstimatorMetadata

Get estimator metadata

impl Estimator for VariationalSparseGaussianProcessRegressor<VsgprTrained>

type Config = GpcConfig

Configuration type for the estimator

type Error = SklearsError

Error type for the estimator

type Float = f64

The numeric type used by this estimator

fn config(&self) -> &Self::Config

Get estimator configuration

fn validate_config(&self) -> Result<(), SklearsError>

Validate estimator configuration with detailed error context

fn check_compatibility( &self, n_samples: usize, n_features: usize, ) -> Result<(), SklearsError>

Check if estimator is compatible with given data dimensions

fn metadata(&self) -> EstimatorMetadata

Get estimator metadata

impl Fit<ArrayBase<ViewRepr<&f64>, Dim<[usize; 2]>>, ArrayBase<ViewRepr<&f64>, Dim<[usize; 1]>>> for VariationalSparseGaussianProcessRegressor<Untrained>

type Fitted = VariationalSparseGaussianProcessRegressor<VsgprTrained>

The fitted model type

fn fit( self, X: &ArrayView2<'_, f64>, y: &ArrayView1<'_, f64>, ) -> SklResult<Self::Fitted>

Fit the model to the provided data with validation

fn fit_with_validation( self, x: &X, y: &Y, _x_val: Option<&X>, _y_val: Option<&Y>, ) -> Result<(Self::Fitted, FitMetrics), SklearsError>

where Self: Sized,

Fit with custom validation and early stopping

impl Predict<ArrayBase<ViewRepr<&f64>, Dim<[usize; 2]>>, ArrayBase<OwnedRepr<f64>, Dim<[usize; 1]>>> for VariationalSparseGaussianProcessRegressor<VsgprTrained>

fn predict(&self, X: &ArrayView2<'_, f64>) -> SklResult<Array1<f64>>

Make predictions on the provided data

fn predict_with_uncertainty( &self, x: &X, ) -> Result<(Output, UncertaintyMeasure), SklearsError>

Make predictions with confidence intervals