Struct GraphSageEmbedder 

pub struct GraphSageEmbedder { /* private fields */ }

GraphSAGE embedder trained on (subject, predicate, object) triple lists.

Implements Hamilton et al. (2017) mean aggregator: for each hop, samples up to K neighbours, computes their mean, concatenates with the node’s own representation, applies W_l, ReLU, and L2-normalisation. Trained via margin ranking loss with sign-SGD and gradient clipping.

Implementations

impl GraphSageEmbedder

pub fn new(config: GraphSageEmbedderConfig) -> Self

Create a new, un-trained embedder.

pub fn fit( &mut self, triples: &[(String, String, String)], ) -> Result<(), EmbeddingError>

Train on (subject_iri, predicate_iri, object_iri) triples. After training, embed_entity works for all seen entities and returns a zero vector for any unseen entity (inductive fallback).

pub fn embed_entity(&self, entity: &str) -> Result<Vec<f64>, EmbeddingError>

Return the embedding for an entity IRI. Unknown entities → zero vector.

pub fn is_trained(&self) -> bool

pub fn num_entities(&self) -> usize

pub fn embedding_dim(&self) -> usize

pub fn relu(x: f64) -> f64

ReLU activation (scalar).

pub fn sample_neighbors<'a>( &self, node_iri: &str, adjacency: &'a HashMap<String, Vec<String>>, ) -> Vec<&'a str>

Sample up to neighbor_sample_k neighbour IRIs using a deterministic LCG.

Trait Implementations

impl Debug for GraphSageEmbedder

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

Formats the value using the given formatter.