Struct GraphVAE 

pub struct GraphVAE { /* private fields */ }

Graph Variational Autoencoder (GraphVAE) Learns a probabilistic latent representation of graphs

Implementations

impl GraphVAE

pub fn new( in_features: usize, hidden_features: usize, latent_dim: usize, beta: f32, use_bias: bool, ) -> Self

Create a new Graph Variational Autoencoder

pub fn encode(&self, graph: &GraphData) -> (Tensor, Tensor)

Encode graph to latent distribution parameters

pub fn reparameterize(&self, mu: &Tensor, logvar: &Tensor) -> Tensor

Reparameterization trick for sampling from latent distribution

pub fn decode(&self, z: &Tensor, num_nodes: usize) -> GraphData

Decode latent representation to graph

pub fn forward(&self, graph: &GraphData) -> (GraphData, Tensor, Tensor)

Forward pass through GraphVAE

pub fn compute_loss( &self, graph: &GraphData, reconstructed: &GraphData, mu: &Tensor, logvar: &Tensor, ) -> f32

Compute VAE loss (reconstruction + KL divergence)

pub fn generate(&self, num_nodes: usize) -> GraphData

Generate new graph from random latent vector

pub fn interpolate( &self, graph1: &GraphData, graph2: &GraphData, alpha: f32, num_nodes: usize, ) -> GraphData

Interpolate between two graphs in latent space

Trait Implementations

impl Debug for GraphVAE

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

Formats the value using the given formatter.

impl GraphLayer for GraphVAE

fn forward(&self, graph: &GraphData) -> GraphData

Forward pass through the layer

fn parameters(&self) -> Vec<Tensor>

Get layer parameters