Module sparse 

Sparse matrix formats, sparse operations and helpers to transform different formats into each other.

Structs

BlockLanczosParams

Parameters for the block thick-restart Lanczos eigensolver.

CompressedSparseData2

Structure to store compressed sparse data with two potential data layers. Think for example in single cell raw counts and normalised counts.

LanczosParams

Parameters for the restarted Lanczos eigensolver.

LanczosResult

Outcome of a restarted Lanczos solve, with the convergence diagnostics.

RitzPairs

Ritz pairs and convergence diagnostics from [lanczos_thick_restart] and block_lanczos_thick_restart.

ScatterPtr

Destination pointer handed to every worker of the parallel scatter.

SparseAxis

Generate structure to store sparse rows or columns

SparseColMoments

One sparse column reduced to what sparse_pairwise_correlations needs.

Enums

CompressedSparseFormat

Type to describe the CompressedSparseFormat

Traits

SparseFromDense

Builds a CompressedSparseData2 from a dense faer matrix.

Functions

block_lanczos_thick_restart

Largest eigenpairs of a symmetric operator via block thick-restarted Lanczos.

compute_largest_eigenpairs_lanczos

Compute the largest eigenpairs of a symmetric sparse matrix.

compute_largest_eigenpairs_lanczos_diag

Compute the largest eigenpairs of a symmetric sparse matrix, with diagnostics.

coo_to_csr

Transform COO stored data into CSR

coo_to_csr_presorted

Optimised COO to CSR - assumes input is already sorted by (row, col)

count_zeroes

Counts the zeroes in a given faer matrix

csr_matmul_csr

Multiply two CSR matrices using sparse accumulators and parallel processing

csr_matmul_dense_block

Apply a CSR operator to a dense row-major block: out = a @ block

csr_matvec

Sparse matrix @ dense vector: M @ v.

csr_sparse_matmul_dense

Sparse CSR @ sparse CSR -> dense

csr_vecmat

Dense vector @ sparse matrix: v @ M.

eliminate_zeros_csr

Remove zeros from sparse matrix

frobenius_norm

Compute Frobenius norm of sparse matrix

frobenius_norm_sq_f64

Squared Frobenius norm of a sparse matrix, in f64.

normalise_csr_columns_l1

Normalises the columns of a CSR matrix to a sum of 1 (L1 norm)

normalise_csr_rows_l1

Normalises the rows of a CSR matrix to a sum of 1 (L1 norm)

parse_compressed_sparse_format

Helper function to parse compressed sparse format

require_csc

Fail unless the matrix is stored as CSC.

require_csr

Fail unless the matrix is stored as CSR.

require_matmul_operands

Check that two CSR matrices can be multiplied, a @ b.

sparse_add_csr

Add two CSR matrices together

sparse_col_means_csc

Calculate the column means for CSC CompressedSparseData2

sparse_col_moments

Reduce one sparse column to its moments without densifying it.

sparse_col_sds_csc

Calculate the column standard deviations for CSC CompressedSparseData2

sparse_multiply_elementwise_csr

Element-wise sparse multiplication

sparse_pairwise_correlations

Pearson correlation of specified column pairs, over the sparsity patterns.

sparse_scalar_multiply_csr

Scalar multiplication of CSR matrix

sparse_subtract_csr

Sparse matrix subtraction

sparse_svd_lanczos

Compute sparse SVD using Lanczos on A^T A or AA^T

thread_chunks

Split 0..n into one contiguous range per rayon worker.

transpose_sparse

Re-express a compressed sparse matrix in the other format (CSC→CSR or CSR→CSC).

transpose_sparse_single_layer

Transpose a compressed sparse matrix (CSC→CSR or CSR→CSC), keeping only one of the two data layers.

undirected_edges_to_csr

Scatter a sorted undirected edge list into a symmetric CSR adjacency

validate_square_csr

Check that a matrix is a structurally sound square CSR and return its order