single_svdlib/
utils.rs

1use rayon::iter::ParallelIterator;
2use nalgebra_sparse::na::{DMatrix, DVector};
3use ndarray::{Array1, Array2};
4use num_traits::{Float, Zero};
5use rayon::prelude::{IntoParallelIterator, IndexedParallelIterator};
6use single_utilities::traits::FloatOpsTS;
7use std::fmt::Debug;
8
9pub fn determine_chunk_size(nrows: usize) -> usize {
10    let num_threads = rayon::current_num_threads();
11
12    let min_rows_per_thread = 16;
13    let desired_chunks_per_thread = 4;
14
15    let target_total_chunks = num_threads * desired_chunks_per_thread;
16    let chunk_size = nrows.div_ceil(target_total_chunks);
17
18    chunk_size.max(min_rows_per_thread)
19}
20
21pub trait SMat<T: Float> {
22    fn nrows(&self) -> usize;
23    fn ncols(&self) -> usize;
24    fn nnz(&self) -> usize;
25    fn svd_opa(&self, x: &[T], y: &mut [T], transposed: bool); // y = A*x
26    fn compute_column_means(&self) -> Vec<T>;
27    fn multiply_with_dense(&self, dense: &DMatrix<T>, result: &mut DMatrix<T>, transpose_self: bool);
28
29    fn multiply_with_dense_centered(&self, dense: &DMatrix<T>, result: &mut DMatrix<T>, transpose_self: bool, means: &DVector<T>);
30}
31
32/// Singular Value Decomposition Components
33///
34/// # Fields
35/// - d:  Dimensionality (rank), the number of rows of both `ut`, `vt` and the length of `s`
36/// - ut: Transpose of left singular vectors, the vectors are the rows of `ut`
37/// - s:  Singular values (length `d`)
38/// - vt: Transpose of right singular vectors, the vectors are the rows of `vt`
39/// - diagnostics: Computational diagnostics
40#[derive(Debug, Clone, PartialEq)]
41pub struct SvdRec<T: Float> {
42    pub d: usize,
43    pub u: Array2<T>,
44    pub s: Array1<T>,
45    pub vt: Array2<T>,
46    pub diagnostics: Diagnostics<T>,
47}
48
49/// Computational Diagnostics
50///
51/// # Fields
52/// - non_zero:  Number of non-zeros in the matrix
53/// - dimensions: Number of dimensions attempted (bounded by matrix shape)
54/// - iterations: Number of iterations attempted (bounded by dimensions and matrix shape)
55/// - transposed:  True if the matrix was transposed internally
56/// - lanczos_steps: Number of Lanczos steps performed
57/// - ritz_values_stabilized: Number of ritz values
58/// - significant_values: Number of significant values discovered
59/// - singular_values: Number of singular values returned
60/// - end_interval: left, right end of interval containing unwanted eigenvalues
61/// - kappa: relative accuracy of ritz values acceptable as eigenvalues
62/// - random_seed: Random seed provided or the seed generated
63#[derive(Debug, Clone, PartialEq)]
64pub struct Diagnostics<T: Float> {
65    pub non_zero: usize,
66    pub dimensions: usize,
67    pub iterations: usize,
68    pub transposed: bool,
69    pub lanczos_steps: usize,
70    pub ritz_values_stabilized: usize,
71    pub significant_values: usize,
72    pub singular_values: usize,
73    pub end_interval: [T; 2],
74    pub kappa: T,
75    pub random_seed: u32,
76}
77
78pub trait SvdFloat: FloatOpsTS {
79    fn eps() -> Self;
80    fn eps34() -> Self;
81    fn compare(a: Self, b: Self) -> bool;
82}
83
84impl SvdFloat for f32 {
85    fn eps() -> Self {
86        f32::EPSILON
87    }
88
89    fn eps34() -> Self {
90        f32::EPSILON.powf(0.75)
91    }
92
93    fn compare(a: Self, b: Self) -> bool {
94        (b - a).abs() < f32::EPSILON
95    }
96}
97
98impl SvdFloat for f64 {
99    fn eps() -> Self {
100        f64::EPSILON
101    }
102
103    fn eps34() -> Self {
104        f64::EPSILON.powf(0.75)
105    }
106
107    fn compare(a: Self, b: Self) -> bool {
108        (b - a).abs() < f64::EPSILON
109    }
110}
single_svdlib/utils.rs

single_svdlib/
utils.rs
