Struct MultiTaskGaussianProcessRegressor 

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

Multi-Task Gaussian Process Regressor

This implementation allows learning multiple related tasks simultaneously by sharing information through a shared latent function while maintaining task-specific variations.

Mathematical Background

For each task t, the model assumes: f_t(x) = w_shared * f_shared(x) + w_task * f_task_t(x)

where the covariance between tasks i and j at points x and x’ is: cov[f_i(x), f_j(x’)] = w_shared² * k_shared(x, x’) + δ_{i,j} * w_task² * k_task(x, x’)

Examples

use sklears_gaussian_process::{MultiTaskGaussianProcessRegressor, kernels::RBF};
use sklears_core::traits::{Fit, Predict};
// SciRS2 Policy - Use scirs2-autograd for ndarray types and operations
use scirs2_core::ndarray::array;

let X1 = array![[1.0], [2.0], [3.0], [4.0]];
let y1 = array![1.0, 4.0, 9.0, 16.0];
let X2 = array![[1.5], [2.5], [3.5], [4.5]];
let y2 = array![2.0, 6.0, 12.0, 20.0];

let shared_kernel = RBF::new(1.0);
let task_kernel = RBF::new(0.5);
let mtgp = MultiTaskGaussianProcessRegressor::new()
    .shared_kernel(Box::new(shared_kernel))
    .task_kernel(Box::new(task_kernel))
    .alpha(1e-6);

let mut mtgp = mtgp.add_task("task1", &X1.view(), &y1.view()).unwrap();
mtgp = mtgp.add_task("task2", &X2.view(), &y2.view()).unwrap();
let fitted = mtgp.fit().unwrap();
let predictions = fitted.predict_task("task1", &X1.view()).unwrap();

Implementations

impl MultiTaskGaussianProcessRegressor<Untrained>

pub fn new() -> Self

Create a new Multi-Task Gaussian Process Regressor

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

Set the shared kernel function

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

Set the task-specific kernel function

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

Set the regularization parameter

pub fn shared_weight(self, weight: f64) -> Self

Set the weight for the shared component

pub fn task_weight(self, weight: f64) -> Self

Set the weight for the task-specific components

pub fn add_task( self, task_name: &str, X: &ArrayView2<'_, f64>, y: &ArrayView1<'_, f64>, ) -> SklResult<Self>

Add a task with its training data

pub fn remove_task(self, task_name: &str) -> Self

Remove a task

pub fn task_names(&self) -> Vec<String>

Get the list of task names

impl MultiTaskGaussianProcessRegressor<Untrained>

pub fn fit(self) -> SklResult<MultiTaskGaussianProcessRegressor<MtgpTrained>>

Fit the multi-task Gaussian process

impl MultiTaskGaussianProcessRegressor<MtgpTrained>

pub fn trained_state(&self) -> &MtgpTrained

Access the trained state

pub fn log_marginal_likelihood_task(&self, task_name: &str) -> Option<f64>

Get the log marginal likelihood for a specific task

pub fn log_marginal_likelihoods(&self) -> &HashMap<String, f64>

Get all log marginal likelihoods

pub fn task_names(&self) -> Vec<String>

Get the list of available tasks

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

Predict for a specific task

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

Get shared and task-specific contributions separately

Trait Implementations

impl<S: Clone> Clone for MultiTaskGaussianProcessRegressor<S>

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

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<S: Debug> Debug for MultiTaskGaussianProcessRegressor<S>

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

Formats the value using the given formatter.

impl Default for MultiTaskGaussianProcessRegressor<Untrained>

fn default() -> Self

Returns the “default value” for a type.

impl Estimator for MultiTaskGaussianProcessRegressor<Untrained>

type Config = MtgpConfig

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 MultiTaskGaussianProcessRegressor<MtgpTrained>

type Config = MtgpConfig

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