Struct PrototypicalDistance 

pub struct PrototypicalDistance { /* private fields */ }

Prototypical distance calculator for few-shot learning.

Computes distances between query examples and class prototypes. Prototypes are computed as the mean of support examples for each class.

Implementations

impl PrototypicalDistance

pub fn euclidean() -> Self

Create with Euclidean distance.

pub fn cosine() -> Self

Create with cosine distance.

pub fn new(metric: DistanceMetric) -> Self

Create with custom distance metric.

pub fn compute_prototypes(&mut self, support: &SupportSet)

Compute prototypes from support set.

Arguments

• support - Support set with labeled examples

pub fn compute_distances( &self, query: &ArrayView1<'_, f64>, ) -> TrainResult<Array1<f64>>

Compute distances from query to all prototypes.

Arguments

• query - Query feature vector

Returns

Distance to each prototype (class)

pub fn predict(&self, query: &ArrayView1<'_, f64>) -> TrainResult<usize>

Predict class for query example.

Arguments

• query - Query feature vector

Returns

Predicted class (nearest prototype)

pub fn predict_proba( &self, query: &ArrayView1<'_, f64>, temperature: f64, ) -> TrainResult<Array1<f64>>

Predict probabilities using softmax over negative distances.

Arguments

• query - Query feature vector
• temperature - Temperature for softmax (default 1.0)

Returns

Probability distribution over classes

Trait Implementations

impl Clone for PrototypicalDistance

fn clone(&self) -> PrototypicalDistance

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PrototypicalDistance

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

Formats the value using the given formatter.