Struct SparseCOO 

pub struct SparseCOO {
    pub indices: Vec<Vec<usize>>,
    pub values: Vec<f32>,
    pub shape: Vec<usize>,
    pub is_coalesced: bool,
}

Sparse tensor in COO (Coordinate) format.

Stores non-zero elements as a list of (index, value) pairs. Efficient for construction but less efficient for arithmetic.

Fields

indices: Vec<Vec<usize>>

Row indices of non-zero elements

values: Vec<f32>

Values of non-zero elements

shape: Vec<usize>

Shape of the tensor

is_coalesced: bool

Whether indices are sorted

Implementations

impl SparseCOO

pub fn new( indices: Vec<Vec<usize>>, values: Vec<f32>, shape: Vec<usize>, ) -> Self

Creates a new sparse COO tensor.

Arguments

• indices - List of index tuples, one per dimension
• values - Non-zero values
• shape - Shape of the tensor

pub fn from_coo_2d( coords: &[(usize, usize)], values: &[f32], shape: &[usize], ) -> Self

Creates from a list of 2D coordinate tuples.

pub fn nnz(&self) -> usize

Returns number of non-zero elements.

pub fn density(&self) -> f32

Returns the density (ratio of non-zeros to total elements).

pub fn shape(&self) -> &[usize]

Returns the shape.

pub fn coalesce(&mut self)

Coalesces the sparse tensor (combines duplicate indices).

pub fn to_dense(&self) -> Tensor<f32>

Converts to dense tensor.

pub fn to_csr(&self) -> SparseCSR

Converts to CSR format (for 2D matrices).

Trait Implementations

impl Clone for SparseCOO

fn clone(&self) -> SparseCOO

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for SparseCOO

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

Formats the value using the given formatter.