Struct SparseInput 

pub struct SparseInput<S, F>
where
    S: BixverseNumeric + Into<F> + Float,
    F: BixverseFloat,
{ /* private fields */ }

Sparse NMF input backed by a matched (CSR, CSC) pair.

Validates non-negativity and finiteness once at construction, then computes and caches ||V||_F^2.

Implementations

impl<S, F> SparseInput<S, F>

where S: BixverseNumeric + BixverseSimd + Into<F> + Float + Clone, F: BixverseFloat + BixverseSimd + Sum,

pub fn from_primary<T, U>( matrix: &CompressedSparseData2<T, U>, ) -> Result<Self, MethodsErrors>

where T: BixverseNumeric + Into<S>, U: BixverseNumeric,

Construct a SparseInput from the primary data layer of matrix.

Casts and validates data, derives the transpose, and builds the matched (CSR, CSC) pair.

Params

• matrix - Source sparse matrix whose data field is used.

Returns

A SparseInput, or MethodsErrors::NmfNonFinite / MethodsErrors::NmfNonNegativeViolated if validation fails.

pub fn from_secondary<T, U>( matrix: &CompressedSparseData2<T, U>, ) -> Result<Self, MethodsErrors>

where T: BixverseNumeric, U: BixverseNumeric + Into<S>,

Construct a SparseInput from the secondary data layer of matrix.

Casts and validates data_2, derives the transpose, and builds the matched (CSR, CSC) pair. Intended for the log1p-normalised layer in the single-cell case.

Params

• matrix - Source sparse matrix whose data_2 field is used.

Returns

A SparseInput, or CoreErrors::Data2NotAvailable if data_2 is absent, or MethodsErrors::NmfNonFinite / MethodsErrors::NmfNonNegativeViolated if validation fails.

pub fn csr(&self) -> &CompressedSparseData2<S>

The CSR layout of V, used for V H^T.

Exposed so a GPU backend can upload the already-validated matched pair rather than rebuilding and revalidating it.

Returns

A reference to the CSR layer.

pub fn csc(&self) -> &CompressedSparseData2<S>

The CSC layout of V, used for W^T V.

Exposed for the same reason as Self::csr.

Returns

A reference to the CSC layer.

Trait Implementations

impl<S, F> Debug for SparseInput<S, F>

where S: BixverseNumeric + Into<F> + Float + Debug, F: BixverseFloat + Debug,

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

Formats the value using the given formatter.

impl<S, F> NmfInput<F> for SparseInput<S, F>

where S: BixverseNumeric + BixverseSimd + Into<F> + Float + Clone + Sync, F: BixverseFloat + BixverseSimd + Send + Sync + Sum,

NmfInput trait implementation for SparseInput

fn shape(&self) -> (usize, usize)

Returns the shape

fn sq_frob(&self) -> F

The squared Frobenius norm ||V||_F^2,

fn wt_v(&self, w: MatRef<'_, F>, out: &mut Mat<F>)

Compute W^T V.

fn v_ht(&self, h: MatRef<'_, F>, out: &mut Mat<F>)

Gram matrix W^T W.

fn top_k_svd(&self, k: usize) -> Result<SvdResults<F>, MethodsErrors>

Top-k truncated SVD of V.