Sparse helpers cover both sorted-index intersection and weighted sparse dot products.
use numkong::{SparseIntersect, SparseDot};
let a_idx = [1_u32, 3, 5, 7];
let b_idx = [3_u32, 4, 5, 8];
let count = u32::sparse_intersection_size(&a_idx, &b_idx).unwrap();
assert_eq!(count, 2); // indices 3 and 5
let a_weights = [1.0_f32, 2.0, 3.0, 4.0];
let b_weights = [5.0_f32, 6.0, 7.0, 8.0];
let dot = u32::sparse_dot(&a_idx, &b_idx, &a_weights, &b_weights).unwrap();
assert!(dot > 0.0); // weighted dot over shared indicesSparseDotComputes sparse dot product between two sorted sparse vectors with weights.3pub trait SparseDot: SizedEach vector consists of sorted indices and corresponding weights. The dot product is computed over the intersection of indices, summing the products of weights.
Error::KernelFailed when no capability of the current
Capabilities has the kernel.
WeightWeight type for this sparse dot product.type Weight;OutputOutput type for this sparse dot product.type Output;sparse_dotComputes sparse dot product.fn sparse_dot(
a_indices: &[Self],
b_indices: &[Self],
a_weights: &[Self::Weight],
b_weights: &[Self::Weight],
) -> Result<Self::Output, Error>Returns the sum of a_weights[i] × b_weights[j] for all pairs where a_indices[i] == b_indices[j]. A sparse vector carries one weight per index, so a weights slice shorter than
its index slice describes no vector at all and yields the default rather than reading past
its end.
SparseIntersectComputes set operations on sorted sparse vectors.2pub trait SparseIntersect: SizedError::KernelFailed when no capability of the current
Capabilities has the kernel.
impl SparseIntersect for u16impl SparseIntersect for u32impl SparseIntersect for u64sparse_intersection_sizeReturns the intersection size between two sorted sparse vectors.fn sparse_intersection_size(a: &[Self], b: &[Self]) -> Result<usize, Error>sparse_intersect_intoComputes intersection and writes matching elements to output buffer.fn sparse_intersect_into(a: &[Self], b: &[Self], result: &mut [Self]) -> Result<usize, Error>Buffer must be at least min(a.len(), b.len()) in size.
Returns the number of elements written, or Error::CapacityExceeded if the buffer
is too small.