Struct DistMult 

pub struct DistMult {
    pub config: KgEmbeddingConfig,
    pub embeddings: Option<KgEmbeddings>,
    /* private fields */
}

DistMult – bilinear diagonal scoring.

Scoring: Σ(hᵢ · rᵢ · tᵢ).
Trained with softplus (logistic) loss and SGD.

Fields

config: KgEmbeddingConfig

embeddings: Option<KgEmbeddings>

Implementations

impl DistMult

pub fn new(config: KgEmbeddingConfig) -> Self

Create a new, untrained DistMult model.

pub fn train( &mut self, triples: &[KgTriple], num_entities: usize, num_relations: usize, ) -> KgResult<TrainingHistory>

Train using negative-sampling and softplus loss.

pub fn score(&self, triple: &KgTriple) -> KgResult<f64>

Score a triple: Σ(hᵢ · rᵢ · tᵢ).

pub fn predict_tail( &self, head: EntityId, relation: RelationId, top_k: usize, ) -> KgResult<Vec<(EntityId, f64)>>

Rank all entities as candidate tails; return top-k.

pub fn predict_head( &self, relation: RelationId, tail: EntityId, top_k: usize, ) -> KgResult<Vec<(EntityId, f64)>>

Rank all entities as candidate heads; return top-k.

Trait Implementations

impl Clone for DistMult

fn clone(&self) -> DistMult

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DistMult

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

Formats the value using the given formatter.

impl KgModel for DistMult

fn score(&self, triple: &KgTriple) -> KgResult<f64>

Score a (head, relation, tail) triple. Higher ⟹ more plausible.

fn predict_tail( &self, head: EntityId, relation: RelationId, top_k: usize, ) -> KgResult<Vec<(EntityId, f64)>>

Rank all entities as possible tails; returns top-k (entity, score) pairs.

fn predict_head( &self, relation: RelationId, tail: EntityId, top_k: usize, ) -> KgResult<Vec<(EntityId, f64)>>

Rank all entities as possible heads; returns top-k (entity, score) pairs.