Struct GraphSageLayer 

pub struct GraphSageLayer {
    pub w_self: Linear,
    pub w_neigh: Linear,
    pub in_dim: usize,
    pub hidden: usize,
}

One GraphSAGE hop: h = σ(W_self x + W_neigh AGG(N(x))).

AGG is inverse-distance weighted mean supplied by the caller. Activation is ReLU.

Fields

w_self: Linear

w_neigh: Linear

in_dim: usize

hidden: usize

Implementations

impl GraphSageLayer

pub fn new(in_dim: usize, hidden: usize, seed: u64) -> Self

pub fn forward(&self, x_self: &[f32], x_agg: &[f32]) -> Vec<f32>

Forward one node given self features and pre-aggregated neighbor features.

pub fn backward( &self, x_self: &[f32], x_agg: &[f32], dy: &[f32], dw_self_w: &mut [f32], dw_self_b: &mut [f32], dw_neigh_w: &mut [f32], dw_neigh_b: &mut [f32], ) -> (Vec<f32>, Vec<f32>)

Backward. Returns (dx_self, dx_agg).

Trait Implementations

impl Clone for GraphSageLayer

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GraphSageLayer

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

Formats the value using the given formatter.