Struct SparseGraph 

pub struct SparseGraph {
    pub edge_list: Vec<(usize, usize)>,
    pub node_features: Option<Tensor>,
    pub edge_weights: Option<Vec<f32>>,
    pub num_nodes: usize,
    pub num_edges: usize,
}

Sparse coordinate (COO) representation of a graph or graph operator.

Each stored entry is a (row, col) coordinate together with an associated weight. This is used both for sparsified adjacency matrices (see SparseGraph::from_dense) and for sparse operators such as the graph Laplacian (see sparse_laplacian).

Fields

edge_list: Vec<(usize, usize)>

Stored (row, col) coordinates of non-zero entries.

node_features: Option<Tensor>

Optional dense node-feature matrix associated with the graph.

edge_weights: Option<Vec<f32>>

Weight associated with each coordinate in edge_list (same length).

num_nodes: usize

Number of nodes the operator is defined over.

num_edges: usize

Number of stored entries (edge_list.len()).

Implementations

impl SparseGraph

pub fn from_dense(adjacency: &Tensor, threshold: f32) -> Result<Self>

Build a sparse graph from a dense adjacency matrix, keeping only entries whose magnitude strictly exceeds threshold.

The matrix is interpreted as a rows x cols tensor; num_nodes is taken from the number of rows. Entries are scanned in row-major order so the resulting edge_list is sorted by (row, col).

pub fn to_edge_index(&self) -> Result<Tensor>

Convert the stored coordinates into a [2, num_edges] edge-index tensor.

The first row holds source indices and the second holds destinations, matching the convention used by GraphData. An empty graph yields a [2, 0] tensor.

pub fn memory_footprint(&self) -> usize

Total memory occupied by this representation, in bytes.

This counts the inline size of the struct itself plus all heap allocations it owns (the coordinate list, the optional node-feature tensor, and the optional weight vector). The value is therefore always strictly positive, even for a graph with no stored edges, because the structure’s own fields still occupy memory.

pub fn density(&self) -> f32

Fraction of possible directed entries that are actually stored.

Returns num_edges / num_nodes^2, or 0.0 for an empty node set.

Trait Implementations

impl Clone for SparseGraph

fn clone(&self) -> SparseGraph

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SparseGraph

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

Formats the value using the given formatter.