Struct PredictorNetwork 

pub struct PredictorNetwork<T: Float + Debug + ScalarOperand + Send + Sync + 'static> { /* private fields */ }

Performance prediction network: fixed random feature map + ridge-fitted heads.

Implementations

impl<T: Float + Debug + ScalarOperand + Send + Sync + 'static> PredictorNetwork<T>

pub fn new(architecture: Vec<usize>, ridge_lambda: T) -> Result<Self>

Build the network.

architecture lists the widths of the input layer and every hidden layer; the last entry is the width of the feature map the linear heads read. ridge_lambda must be positive so the normal equations stay positive-definite (and hence Cholesky-solvable) even with fewer samples than features.

Errors

Returns Err when the architecture has fewer than two layers, any width is zero, the input width disagrees with PREDICTION_FEATURE_COUNT, or ridge_lambda <= 0.

pub fn is_trained(&self) -> bool

Whether the heads have been fitted.

pub fn architecture(&self) -> &[usize]

Layer widths.

pub fn feature_width(&self) -> usize

Width of the feature map the heads read.

pub fn parameter_count(&self) -> usize

Total parameter count (fixed map + learned heads).

pub fn fit(&mut self, samples: &[PredictorSample]) -> Result<PredictorFitReport>

Fit both heads by ridge regression on samples.

Solves (ΦᵗΦ + λI) w = Φᵗ y exactly via Cholesky. Returns the training RMSE of each head.

Errors

Returns Err when samples is empty or the normal-equation matrix is not positive-definite (which cannot happen for λ > 0, but is checked rather than assumed).

pub fn predict(&self, features: &PredictionFeatures) -> Result<PredictorOutput>

Predict from a feature vector.

While untrained the means are 0 and the uncertainty is 1 with confidence 0 — an explicit “no information” answer rather than a fabricated constant.

Trait Implementations

impl<T: Clone + Float + Debug + ScalarOperand + Send + Sync + 'static> Clone for PredictorNetwork<T>

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl<T: Debug + Float + Debug + ScalarOperand + Send + Sync + 'static> Debug for PredictorNetwork<T>

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

Formats the value using the given formatter.