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
|
% GET [STK internal]
%
% CALL: value = get (kreq, propname)
%
% implements 'get' for stk_kreq_qr objects.
% Copyright Notice
%
% Copyright (C) 2015, 2017 CentraleSupelec
% Copyright (C) 2013, 2014 SUPELEC
%
% Author: Julien Bect <julien.bect@centralesupelec.fr>
% Emmanuel Vazquez <emmanuel.vazquez@centralesupelec.fr>
% Copying Permission Statement
%
% This file is part of
%
% STK: a Small (Matlab/Octave) Toolbox for Kriging
% (http://sourceforge.net/projects/kriging)
%
% STK is free software: you can redistribute it and/or modify it under
% the terms of the GNU General Public License as published by the Free
% Software Foundation, either version 3 of the License, or (at your
% option) any later version.
%
% STK is distributed in the hope that it will be useful, but WITHOUT
% ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
% or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public
% License for more details.
%
% You should have received a copy of the GNU General Public License
% along with STK. If not, see <http://www.gnu.org/licenses/>.
function value = get (kreq, propname)
switch propname
case {'n', 'r', 'P_scaling', 'LS_Q', 'LS_R', 'RS', 'lambda_mu'}
value = kreq.(propname);
case 'lambda'
value = kreq.lambda_mu(1:kreq.n, :);
case 'mu'
value = kreq.lambda_mu((kreq.n + 1):end, :);
case 'delta_var'
value = dot (kreq.lambda_mu, kreq.RS, 1)';
case 'log_abs_det_kriging_matrix'
% LS_Q has +/- 1 determinant, so we only need to care about LS_R
value = sum (log (abs (diag (kreq.LS_R))));
case 'log_det_covariance_matrix_a'
n = kreq.n;
Q11 = kreq.LS_Q(1:n, 1:n);
R11 = kreq.LS_R(1:n, 1:n);
value = log (det (Q11 * R11)); % K = Q11 * R11
% note: this method seems much less accurate than the others,
% we strongly recommend to avoid using it
case 'log_det_covariance_matrix_b'
n = kreq.n;
Q11 = kreq.LS_Q(1:n, 1:n);
R11 = kreq.LS_R(1:n, 1:n);
Kchol = stk_cholcov (Q11 * R11); % K = Q11 * R11
value = 2 * sum (log (diag (Kchol)));
case 'log_det_covariance_matrix_c'
% note: very efficient (but not the best) and quite simple to understand
n = kreq.n;
Q11 = kreq.LS_Q(1:n, 1:n);
diag_R = diag (kreq.LS_R);
logdet1 = log (abs (det (Q11)));
logdet2 = sum (log (abs (diag_R(1:n))));
value = logdet1 + logdet2;
case {'log_det_covariance_matrix', 'log_det_covariance_matrix_d'}
% note: this is the fastest of all five solutions for large matrices
% (for small matrices, the difference is negligible)... but also
% the trickiest !
n = kreq.n;
QL = kreq.LS_Q((n + 1):end, :); % lowerpart of LS_Q
diag_R = diag (kreq.LS_R);
logdet1 = sum (log (abs (diag_R)));
T = kreq.LS_R \ (QL');
logdet2 = log (abs (det (T((n + 1):end, :))));
value = logdet1 + logdet2;
case 'log_det_covariance_matrix_e'
% note: the most time consuming of all five solutions (blame qrdelete),
% we strongly recommend to avoid using it
n = kreq.n;
r = kreq.r;
Q = kreq.LS_Q;
R = kreq.LS_R;
for i = 0:(r - 1)
[Q, R] = qrdelete (Q, R, n + r - i, 'row');
[Q, R] = qrdelete (Q, R, n + r - i, 'col');
end
value = sum (log (abs (diag (R))));
otherwise
error ('There is no property called ''%s''.', propname);
end % switch
end % function
|
