Struct TransE 

pub struct TransE {
    pub entity_embeddings: Vec<Vec<f64>>,
    pub relation_embeddings: Vec<Vec<f64>>,
    pub dim: usize,
    pub norm_order: u32,
}

TransE knowledge graph embedding model.

Score: −‖ h + r − t ‖_p (negated L-p distance, higher = more likely).

Training uses margin-based loss with negative sampling.

Fields

entity_embeddings: Vec<Vec<f64>>

Entity embedding table [n_entities, dim]

relation_embeddings: Vec<Vec<f64>>

Relation embedding table [n_relations, dim]

dim: usize

Embedding dimension

norm_order: u32

Norm order (1 or 2)

Implementations

impl TransE

pub fn new(n_entities: usize, n_relations: usize, dim: usize) -> Result<Self>

Create a TransE model with random initialisation.

Arguments

• n_entities – Number of entities.
• n_relations – Number of relation types.
• dim – Embedding dimension.

pub fn score(&self, h: usize, r: usize, t: usize) -> Result<f64>

Score a single triple: −‖ h + r − t ‖ (higher = more plausible).

pub fn predict_tails(&self, h: usize, r: usize, k: usize) -> Result<Vec<usize>>

Return the top-k entity indices most likely to complete (h, r, ?).

pub fn link_prediction_score(&self, h: usize, r: usize, t: usize) -> Result<f64>

Link prediction score for a triple (h, r, t).

pub fn train_epoch(&mut self, dataset: &KGDataset, lr: f64, margin: f64) -> f64

Train for one epoch using stochastic gradient descent and negative sampling. Returns the total margin-ranking loss.

Arguments

• dataset – Training triples.
• lr – Learning rate.
• margin – Margin γ for the margin-ranking loss.

Trait Implementations

impl Clone for TransE

fn clone(&self) -> TransE

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TransE

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

Formats the value using the given formatter.