faer 0.24.0

linear algebra library
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146

use crate::assert;
use crate::internal_prelude::*;
use linalg::matmul::triangular::BlockStructure;
pub fn reconstruct_scratch<I: Index, T: ComplexField>(
	dim: usize,
	par: Par,
) -> StackReq {
	_ = par;
	temp_mat_scratch::<T>(dim, dim)
}
#[track_caller]
pub fn reconstruct<I: Index, T: ComplexField>(
	out: MatMut<'_, T>,
	L: MatRef<'_, T>,
	diagonal: DiagRef<'_, T>,
	subdiagonal: DiagRef<'_, T>,
	perm: PermRef<'_, I>,
	par: Par,
	stack: &mut MemStack,
) {
	let n = L.nrows();
	assert!(all(
		out.nrows() == n,
		out.ncols() == n,
		L.nrows() == n,
		L.ncols() == n,
		diagonal.dim() == n,
		subdiagonal.dim() == n,
		perm.len() == n,
	));
	let (mut tmp, _) = unsafe { temp_mat_uninit::<T, _, _>(n, n, stack) };
	let mut tmp = tmp.as_mat_mut();
	let mut out = out;
	let s = subdiagonal;
	out.fill(zero());
	out.rb_mut().diagonal_mut().fill(one());
	out.copy_from_strict_triangular_lower(L);
	let mut j = 0;
	while j < n {
		if s[j] == zero() {
			let d = &L[(j, j)].real();
			for i in 0..n {
				out[(i, j)] = out[(i, j)].mul_real(d);
			}
			j += 1;
		} else {
			let akp1k = &s[j].copy();
			let ak = &L[(j, j)].real();
			let akp1 = &L[(j + 1, j + 1)].real();
			for i in 0..n {
				let xk = &out[(i, j)];
				let xkp1 = &out[(i, j + 1)];
				(out[(i, j)], out[(i, j + 1)]) = (
					xk.mul_real(ak) + xkp1 * akp1k,
					xkp1.mul_real(akp1) + xk * akp1k.conj(),
				);
			}
			j += 2;
		}
	}
	linalg::matmul::triangular::matmul(
		tmp.rb_mut(),
		BlockStructure::TriangularLower,
		Accum::Replace,
		L,
		BlockStructure::UnitTriangularLower,
		out.rb().adjoint(),
		BlockStructure::Rectangular,
		one(),
		par,
	);
	let perm_inv = perm.arrays().1;
	for j in 0..n {
		let pj = perm_inv[j].zx();
		for i in j..n {
			let pi = perm_inv[i].zx();
			out[(i, j)] = if pi >= pj {
				tmp[(pi, pj)].copy()
			} else {
				tmp[(pj, pi)].conj()
			};
		}
	}
	for j in 0..n {
		out[(j, j)] = out[(j, j)].as_real();
	}
}
#[cfg(test)]
mod tests {
	use super::*;
	use crate::assert;
	use crate::stats::prelude::*;
	use crate::utils::approx::*;
	use dyn_stack::MemBuffer;
	use linalg::cholesky::lblt::*;
	#[test]
	fn test_reconstruct() {
		let rng = &mut StdRng::seed_from_u64(0);
		let n = 50;
		let A = CwiseMatDistribution {
			nrows: n,
			ncols: n,
			dist: ComplexDistribution::new(StandardNormal, StandardNormal),
		}
		.rand::<Mat<c64>>(rng);
		let A = &A + A.adjoint();
		let mut LB = A.to_owned();
		let mut subdiag = Diag::zeros(n);
		let perm_fwd = &mut *vec![0usize; n];
		let perm_bwd = &mut *vec![0usize; n];
		let (_, perm) = factor::cholesky_in_place(
			LB.as_mut(),
			subdiag.as_mut(),
			perm_fwd,
			perm_bwd,
			Par::Seq,
			MemStack::new(&mut {
				MemBuffer::new(factor::cholesky_in_place_scratch::<usize, c64>(
					n,
					Par::Seq,
					default(),
				))
			}),
			default(),
		);
		let approx_eq = CwiseMat(ApproxEq::eps() * (n as f64));
		let mut A_rec = Mat::zeros(n, n);
		reconstruct::reconstruct(
			A_rec.as_mut(),
			LB.as_ref(),
			LB.diagonal(),
			subdiag.as_ref(),
			perm,
			Par::Seq,
			MemStack::new(&mut MemBuffer::new(
				reconstruct::reconstruct_scratch::<usize, c64>(n, Par::Seq),
			)),
		);
		for j in 0..n {
			for i in 0..j {
				A_rec[(i, j)] = A_rec[(j, i)].conj();
			}
		}
		assert!(A_rec ~ A);
	}
}