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
|
function [U, R, V, r] = cod (A, tol)
%COD complete orthogonal decomposition of a full matrix A = U*R*V'
%
% [U, R, V, r] = cod (A)
% [U, R, V, r] = cod (A, tol)
%
% The full m-by-n matrix A is factorized into U*R*V' where R is r-by-r and
% upper triangular and where r is the estimated rank of A. The diagonal
% entries of R have magnitude greater than tol. The default tol of
% 20*(m+n)*eps(max(diag(R))) is used if tol is not present or if tol < 0.
% Use COD_SPARSE for sparse matrices.
%
% COD provides a 'rank-sized' economy factorization, where R is r-by-r, U is
% m-by-r and V is n-by-r. U and V are matrices with orthonormal columns. That
% is, U'*U = V'*V = eye (r). If A is rank deficient, then U and V are full
% matrices. If A has full rank, either U or V are sparse permutation matrices
% (V if m >= n, U if m < n). QR with column 2-norm pivoting is used on A
% if m >= n, or on A' if m < n, and thus abs(diag(R)) is decreasing if m >= n
% and increasing if m < n.
%
% If condest(R) is high (> 1e12 or so), then the estimated rank of A
% might be incorrect. Try increasing tol in that case, which will make R
% better conditioned and reduce the estimated rank.
%
% Example:
%
% A = magic (4), [U, R, V] = cod (A), norm (A - U*R*V')
% A = rand (4,3), [U, R, V] = cod (A), norm (A - U*R*V')
% A = rand (3,4), [U, R, V] = cod (A), norm (A - U*R*V')
%
% See also qr, svd, rq, spqr, cod_sparse.
% Factorize, Copyright (c) 2011-2012, Timothy A Davis. All Rights Reserved.
% SPDX-License-Identifier: BSD-3-clause
if (issparse (A))
error ('FACTORIZE:cod:sparse', ...
'COD is not designed for sparse matrices. Use COD_SPARSE instead.') ;
end
[m, n] = size (A) ;
if (nargin < 2)
tol = -1 ; % default tol will be used
end
if (m >= n)
% factorize A into U*R*V' with column 2-norm pivoting
[U, R, V] = qr (A, 0) ; % economy U*R = A(V,:) with column pivoting V
V = sparse (V, 1:n, 1) ; % R n-by-n and triu, U m-by-n, V n-by-n
% find a rough estimate of the rank of A
r = rank_est (R, m, n, tol) ;
if (r < n)
% A is rank deficient. R upper trapezoidal with R(r+1:end,:) tiny
[R, Q] = rq (R, r, n) ; % RQ factorization, R now upper triangular
U = U (:, 1:r) ; % discard all but the first r columns of U
V = V * Q' ; % merge Q and V
% R is now r-by-r, U is m-by-r, and V is n-by-r.
end
else
% compute the cod of A' and permute the result
[V, R, U, r] = cod (A', tol) ;
U = U (:, end:-1:1) ;
R = R (end:-1:1, end:-1:1)' ;
V = V (:, end:-1:1) ;
end
|
