Struct StructuredPerceptron 

pub struct StructuredPerceptron {
    pub n_labels: usize,
    pub n_features: usize,
    pub emissions: Vec<f64>,
    pub transitions: Vec<f64>,
}

A linear-chain structured perceptron tagger.

Parameter layout (identical to LinearChainCrf):

• emissions[label*n_features + k] — weight for feature k under label.
• transitions[prev*n_labels + cur] — score of the prev → cur bigram.

Fields

n_labels: usize

Number of output labels.

n_features: usize

Number of (real-valued) emission features per position.

emissions: Vec<f64>

Emission weights, length n_labels * n_features.

transitions: Vec<f64>

Transition weights, length n_labels * n_labels.

Implementations

impl StructuredPerceptron

pub fn zeros(n_labels: usize, n_features: usize) -> SeqResult<Self>

Create a zero-initialised perceptron.

Errors

SeqError::InvalidConfiguration if n_labels == 0 or n_features == 0.

pub fn param_count(&self) -> usize

Total number of parameters.

pub fn decode(&self, x: &[f64]) -> SeqResult<Vec<usize>>

Viterbi decode the highest-scoring label sequence for feature matrix x (T × n_features, row-major).

Errors

• SeqError::EmptyInput — if x is empty.
• SeqError::ShapeMismatch — if x.len() is not a multiple of n_features.

pub fn sequence_score(&self, x: &[f64], y: &[usize]) -> SeqResult<f64>

Total linear score of a full label sequence y under x.

Errors

• SeqError::EmptyInput — if y is empty.
• SeqError::ShapeMismatch — if x.len() ≠ y.len() * n_features.
• SeqError::IndexOutOfBounds — if any label ≥ n_labels.

pub fn update( &mut self, x: &[f64], gold: &[usize], pred: &[usize], ) -> SeqResult<usize>

Apply the perceptron correction θ ← θ + φ(gold) − φ(pred) in place, returning the number of positions where gold and pred differ.

Both label sequences must have length T = x.len() / n_features.

Errors

• SeqError::LengthMismatch — if gold.len() ≠ pred.len().
• SeqError::ShapeMismatch — if shapes are inconsistent.
• SeqError::IndexOutOfBounds — if any label is out of range.

Trait Implementations

impl Clone for StructuredPerceptron

fn clone(&self) -> StructuredPerceptron

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for StructuredPerceptron

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

Formats the value using the given formatter.