Struct EmbeddingModel 

pub struct EmbeddingModel<N: Node> {
    pub embeddings: HashMap<N, Embedding>,
    pub context_embeddings: HashMap<N, Embedding>,
    pub dimensions: usize,
}

Graph embedding model

Fields

embeddings: HashMap<N, Embedding>

Node embeddings (input vectors)

context_embeddings: HashMap<N, Embedding>

Context embeddings (output vectors) for skip-gram

dimensions: usize

Dimensionality of embeddings

Implementations

impl<N: Node> EmbeddingModel<N>

pub fn new(dimensions: usize) -> Self

Create a new embedding model

pub fn get_embedding(&self, node: &N) -> Option<&Embedding>

Get embedding for a node

pub fn set_embedding(&mut self, node: N, embedding: Embedding) -> Result<()>

Set embedding for a node

pub fn initialize_random<E, Ix>( &mut self, graph: &Graph<N, E, Ix>, rng: &mut impl Rng, )

where N: Clone + Debug, E: EdgeWeight, Ix: IndexType,

Initialize random embeddings for all nodes

pub fn initialize_random_digraph<E, Ix>( &mut self, graph: &DiGraph<N, E, Ix>, rng: &mut impl Rng, )

where N: Clone + Debug, E: EdgeWeight, Ix: IndexType,

Initialize random embeddings for directed graph

pub fn most_similar(&self, node: &N, k: usize) -> Result<Vec<(N, f64)>>

where N: Clone,

Find k most similar nodes to a given node

pub fn generate_context_pairs( walks: &[RandomWalk<N>], window_size: usize, ) -> Vec<ContextPair<N>>

where N: Clone,

Generate context pairs from random walks

pub fn train_skip_gram( &mut self, pairs: &[ContextPair<N>], negative_sampler: &NegativeSampler<N>, learning_rate: f64, negative_samples: usize, rng: &mut impl Rng, ) -> Result<()>

Train skip-gram model on context pairs with negative sampling

Trait Implementations

impl<N: Debug + Node> Debug for EmbeddingModel<N>

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

Formats the value using the given formatter.