Struct CentralityEncoding 

pub struct CentralityEncoding {
    pub in_degree_embed: Array2<f64>,
    pub out_degree_embed: Array2<f64>,
    pub max_degree: usize,
    pub hidden_dim: usize,
}

Learnable centrality encoding based on node in-degree and out-degree.

Adds z_in[deg_in(v)] + z_out[deg_out(v)] to each node embedding, where z_in and z_out are learnable embedding tables indexed by degree.

Fields

in_degree_embed: Array2<f64>

Embedding table for in-degree: [max_degree + 1, hidden_dim]

out_degree_embed: Array2<f64>

Embedding table for out-degree: [max_degree + 1, hidden_dim]

max_degree: usize

Maximum degree supported

hidden_dim: usize

Hidden dimension

Implementations

impl CentralityEncoding

pub fn new(max_degree: usize, hidden_dim: usize) -> Self

Create a new centrality encoding module.

Arguments

• max_degree - Maximum node degree to encode (higher degrees are clamped)
• hidden_dim - Dimension of the centrality embedding

pub fn compute_degrees(&self, adj: &CsrMatrix) -> (Vec<usize>, Vec<usize>)

Compute in-degree and out-degree for each node from adjacency.

pub fn forward( &self, features: &Array2<f64>, adj: &CsrMatrix, ) -> Result<Array2<f64>>

Encode centrality information and add to node features.

Arguments

• features - Node features [n_nodes, hidden_dim]
• adj - Adjacency matrix

Returns

Updated features with centrality encoding added

Trait Implementations

impl Clone for CentralityEncoding

fn clone(&self) -> CentralityEncoding

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CentralityEncoding

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

Formats the value using the given formatter.