Struct GraphSAGELayer 

pub struct GraphSAGELayer {
    pub in_dim: usize,
    pub out_dim: usize,
    /* private fields */
}

A single GraphSAGE layer: aggregate neighbors then combine with self.

For each node v: agg = AGGREGATE({ h_u | u ∈ N(v) }) h_v = σ( W · CONCAT(h_v, agg) )

Fields

in_dim: usize

out_dim: usize

Implementations

impl GraphSAGELayer

pub fn new( in_dim: usize, out_dim: usize, num_samples: usize, kind: &AggregatorKind, rng: &mut Lcg, ) -> Result<Self>

Build a new layer.

in_dim - input feature dimension for this layer
out_dim - output embedding dimension
num_samples - neighborhood sample size
kind - aggregator type
rng - seeded RNG for weight initialization

pub fn forward( &self, graph: &Graph, current_embeddings: &[Vec<f64>], rng: &mut Lcg, ) -> Vec<Vec<f64>>

Forward pass: compute new embeddings for all nodes.

current_embeddings[v] = current feature vector for node v.

Trait Implementations

impl Debug for GraphSAGELayer

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

Formats the value using the given formatter.