Struct SpectralEmbedding 

pub struct SpectralEmbedding<N: Node> { /* private fields */ }

Spectral embedding of a graph

Computes a low-dimensional embedding by finding the smallest eigenvectors of the graph Laplacian matrix using the power iteration / inverse iteration method.

Implementations

impl<N: Node + Debug> SpectralEmbedding<N>

pub fn new(config: SpectralEmbeddingConfig) -> Self

Create a new spectral embedding with the given configuration

pub fn fit<E, Ix>(&mut self, graph: &Graph<N, E, Ix>) -> Result<()>

where N: Clone, E: EdgeWeight + Into<f64> + Zero + One + Copy, Ix: IndexType,

Fit the spectral embedding to a graph

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

Get the embedding for a specific node

pub fn embeddings(&self) -> Result<HashMap<N, Embedding>>

where N: Clone,

Get all embeddings as a HashMap

pub fn embedding_matrix(&self) -> Result<&Array2<f64>>

Get the embedding matrix (n x d)

pub fn eigenvalues(&self) -> Result<&Array1<f64>>

Get the eigenvalues corresponding to the embedding dimensions

pub fn dimensions(&self) -> usize

Get the embedding dimension

pub fn pairwise_distances(&self) -> Result<Array2<f64>>

Compute pairwise distances between all embedded nodes

pub fn compute_stress<E, Ix>(&self, graph: &Graph<N, E, Ix>) -> Result<f64>

where N: Clone, E: EdgeWeight + Into<f64> + Clone, Ix: IndexType,

Compute the embedding quality metric (stress) Lower stress means the embedding better preserves graph distances