Struct AdversarialMultiTaskNetwork 

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

Adversarial Multi-Task Network with feature disentanglement

This network implements adversarial multi-task learning where a task discriminator is trained to predict which task shared features come from, while the shared feature extractor is trained adversarially to fool the discriminator. This ensures that shared representations contain only task-invariant information.

Architecture

The network consists of:

• Shared layers: Learn task-invariant representations
• Private layers: Learn task-specific representations per task
• Task discriminator: Tries to predict task from shared features
• Gradient reversal: Adversarial training mechanism

Examples

use sklears_multioutput::adversarial::{AdversarialMultiTaskNetwork, AdversarialStrategy};
use sklears_core::traits::{Predict, Fit};
// Use SciRS2-Core for arrays and random number generation (SciRS2 Policy)
use scirs2_core::ndarray::array;
use std::collections::HashMap;

let X = array![[1.0, 2.0], [2.0, 3.0], [3.0, 1.0], [4.0, 4.0]];
let mut tasks = HashMap::new();
tasks.insert("task1".to_string(), array![[0.5], [1.0], [1.5], [2.0]]);
tasks.insert("task2".to_string(), array![[1.0], [0.0], [1.0], [0.0]]);

let adv_net = AdversarialMultiTaskNetwork::new()
    .shared_layers(vec![20, 10])
    .private_layers(vec![8])
    .task_outputs(&[("task1", 1), ("task2", 1)])
    .adversarial_strategy(AdversarialStrategy::GradientReversal)
    .adversarial_weight(0.1)
    .orthogonality_weight(0.01)
    .random_state(Some(42));

Implementations

impl AdversarialMultiTaskNetwork<Untrained>

pub fn new() -> Self

Create a new AdversarialMultiTaskNetwork

pub fn shared_layers(self, sizes: Vec<usize>) -> Self

Set shared layer sizes

pub fn private_layers(self, sizes: Vec<usize>) -> Self

Set private layer sizes

pub fn task_outputs(self, tasks: &[(&str, usize)]) -> Self

Configure task outputs

pub fn adversarial_strategy(self, strategy: AdversarialStrategy) -> Self

Set adversarial strategy

pub fn adversarial_weight(self, weight: Float) -> Self

Set adversarial weight

pub fn orthogonality_weight(self, weight: Float) -> Self

Set orthogonality weight

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

Set learning rate

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

Set maximum iterations

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

Set random state

impl AdversarialMultiTaskNetwork<AdversarialMultiTaskNetworkTrained>

pub fn task_loss_curves(&self) -> &HashMap<String, Vec<Float>>

Get task loss curves

pub fn adversarial_loss_curve(&self) -> &[Float]

Get adversarial loss curve

pub fn orthogonality_loss_curve(&self) -> &[Float]

Get orthogonality loss curve

pub fn combined_loss_curve(&self) -> &[Float]

Get combined loss curve

pub fn discriminator_accuracy_curve(&self) -> &[Float]

Get discriminator accuracy curve

pub fn n_iter(&self) -> usize

Get training iterations

Trait Implementations

impl<S: Clone> Clone for AdversarialMultiTaskNetwork<S>

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

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<S: Debug> Debug for AdversarialMultiTaskNetwork<S>

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

Formats the value using the given formatter.

impl Default for AdversarialMultiTaskNetwork<Untrained>

fn default() -> Self

Returns the “default value” for a type.

impl Estimator for AdversarialMultiTaskNetwork<Untrained>

type Config = AdversarialConfig

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]>>, HashMap<String, ArrayBase<OwnedRepr<f64>, Dim<[usize; 2]>>>> for AdversarialMultiTaskNetwork<Untrained>

type Fitted = AdversarialMultiTaskNetwork<AdversarialMultiTaskNetworkTrained>

The fitted model type

fn fit( self, x: &ArrayView2<'_, Float>, y: &HashMap<String, Array2<Float>>, ) -> 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]>>, HashMap<String, ArrayBase<OwnedRepr<f64>, Dim<[usize; 2]>>>> for AdversarialMultiTaskNetwork<AdversarialMultiTaskNetworkTrained>

fn predict( &self, X: &ArrayView2<'_, Float>, ) -> SklResult<HashMap<String, Array2<Float>>>

Make predictions on the provided data

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

Make predictions with confidence intervals