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
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
|
/*===========================================================================
Copyright (C) 2007-2020 Yves Renard, Julien Pommier.
This file is a part of GetFEM
GetFEM is free software; you can redistribute it and/or modify it
under the terms of the GNU Lesser General Public License as published
by the Free Software Foundation; either version 3 of the License, or
(at your option) any later version along with the GCC Runtime Library
Exception either version 3.1 or (at your option) any later version.
This program 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 Lesser General Public
License and GCC Runtime Library Exception for more details.
You should have received a copy of the GNU Lesser General Public License
along with this program; if not, write to the Free Software Foundation,
Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA.
===========================================================================*/
// RECTANGULAR_MATRIX_PARAM
// SQUARED_MATRIX_PARAM
// ENDPARAM;
#include "gmm/gmm_kernel.h"
#include "gmm/gmm_dense_lu.h"
#include "gmm/gmm_dense_qr.h"
#include "gmm/gmm_condition_number.h"
using std::endl; using std::cout; using std::cerr;
using std::ends; using std::cin;
using gmm::size_type;
bool print_debug = false;
// template <typename MAT, typename T> void print_for_matlab(const MAT &m, T) {
// cout.precision(16);
// cout << "[ ";
// for (size_type i = 0; i < gmm::mat_nrows(m); ++i) {
// for (size_type j = 0; j < gmm::mat_ncols(m); ++j) cout << " " << m(i,j);
// if (i != gmm::mat_nrows(m)-1) cout << " ; \n";
// }
// cout << " ]" << endl;
// }
// template <typename MAT, typename T> void print_for_matlab(const MAT &m,
// std::complex<T>) {
// cout.precision(16);
// cout << "[ ";
// for (size_type i = 0; i < gmm::mat_nrows(m); ++i) {
// for (size_type j = 0; j < gmm::mat_ncols(m); ++j)
// cout << " (" << m(i,j).real() << "+" << m(i,j).imag() << "*i)" ;
// if (i != gmm::mat_nrows(m)-1) cout << " ; \n";
// }
// cout << " ]" << endl;
// }
// template <typename MAT> inline void print_for_matlab(const MAT &m)
// { print_for_matlab(m, gmm::linalg_traits<MAT>::value_type()); }
template <typename T> inline T real_or_complex(double a, double, T)
{ return T(a); }
template <typename T> inline
std::complex<T> real_or_complex(double a, double b, std::complex<T>) {
typedef typename gmm::number_traits<T>::magnitude_type R;
return std::complex<T>(R(a), R(b));
}
template <typename T> struct cmp_eval {
bool operator()(T a, T b) {
typedef typename gmm::number_traits<T>::magnitude_type R;
// R prec = gmm::default_tol(R());
R dr = gmm::real(a) - gmm::real(b);
R di = gmm::imag(a) - gmm::imag(b);
if (gmm::abs(dr) > gmm::abs(di)) return (dr<R(0)); else return (di<R(0));
}
};
template <typename T> void sort_eval(std::vector<T> &v) {
std::sort(v.begin(), v.end(), cmp_eval<T>());
}
template <typename MAT1, typename MAT2>
bool test_procedure(const MAT1 &m1_, const MAT2 &m2_) {
MAT1 &m1 = const_cast<MAT1 &>(m1_);
MAT2 &m2 = const_cast<MAT2 &>(m2_);
typedef typename gmm::linalg_traits<MAT1>::value_type T;
typedef typename gmm::number_traits<T>::magnitude_type R;
R prec = gmm::default_tol(R());
R error;
static size_type nb_iter(0);
++nb_iter;
// gmm::qr_factor(A, Q, R) is tested in test_gmm_mult.C
//
// test for gmm::qr_factor(A), apply_house_right and apply_house_left
//
size_type m = gmm::mat_nrows(m1), n = gmm::mat_ncols(m1);
size_type k = size_type(rand() % 50);
if (print_debug) {
static int nexpe = 0;
cout << "Begin experiment " << ++nexpe << "\n\nwith " << m1 << "\n\n";
gmm::set_warning_level(3);
}
gmm::dense_matrix<T> dm1(m, n);
gmm::copy(m1, dm1);
if (m >= n) {
gmm::dense_matrix<T> q(k,m), qaux(k,m), q2(m,k), dm1aux(k,n), m1aux(k,n);
gmm::fill_random(q); gmm::copy(q, qaux);
gmm::mult(q, m1, m1aux);
gmm::qr_factor(dm1);
gmm::copy(dm1, m1);
gmm::apply_house_right(dm1, q);
for (size_type j = 0; j < n; ++j)
for (size_type i = j+1; i < m; ++i)
dm1(i, j) = T(0);
gmm::mult(q, dm1, dm1aux);
gmm::add(gmm::scaled(m1aux, T(-1)), dm1aux);
error = gmm::mat_euclidean_norm(dm1aux);
if (!(error <= prec * R(10000)))
GMM_ASSERT1(false, "Error too large: " << error);
gmm::copy(gmm::identity_matrix(), q);
gmm::apply_house_right(m1, q);
size_type min_km = std::min(k, m);
gmm::dense_matrix<T> a(min_km, min_km), b(min_km, min_km);
gmm::copy(gmm::identity_matrix(), b);
if (k > m) gmm::mult(gmm::conjugated(q), q, a);
else gmm::mult(q, gmm::conjugated(q), a);
gmm::add(gmm::scaled(b, T(-1)), a);
error = gmm::mat_euclidean_norm(a);
if (!(error <= prec * R(10000)))
GMM_ASSERT1(false, "Error too large: " << error);
gmm::copy(gmm::conjugated(qaux), q2);
gmm::apply_house_left(m1, q2);
gmm::mult(gmm::conjugated(q2), dm1, dm1aux);
gmm::add(gmm::scaled(m1aux, T(-1)), dm1aux);
error = gmm::mat_euclidean_norm(dm1aux);
if (!(error <= prec * R(10000)))
GMM_ASSERT1(false, "Error too large: " << error);
}
else {
gmm::dense_matrix<T> q(k,n), qaux(k,n), q2(n,k), dm1aux(k,m), m1aux(k,m);
gmm::fill_random(q); gmm::copy(q, qaux);
gmm::mult(q, gmm::transposed(m1), m1aux);
gmm::qr_factor(gmm::transposed(dm1));
gmm::copy(dm1, m1);
gmm::apply_house_right(gmm::transposed(dm1), q);
for (size_type i = 0; i < m; ++i)
for (size_type j = i+1; j < n; ++j)
dm1(i, j) = T(0);
gmm::mult(q, gmm::transposed(dm1), dm1aux);
gmm::add(gmm::scaled(m1aux, T(-1)), dm1aux);
error = gmm::mat_euclidean_norm(dm1aux);
if (!(error <= prec * R(10000)))
GMM_ASSERT1(false, "Error too large: " << error);
gmm::copy(gmm::identity_matrix(), q);
gmm::apply_house_right(gmm::transposed(m1), q);
size_type min_km = std::min(k, n);
gmm::dense_matrix<T> a(min_km, min_km), b(min_km, min_km);
gmm::copy(gmm::identity_matrix(), b);
if (k > n) gmm::mult(gmm::conjugated(q), q, a);
else gmm::mult(q, gmm::conjugated(q), a);
gmm::add(gmm::scaled(b, T(-1)), a);
error = gmm::mat_euclidean_norm(a);
if (!(error <= prec * R(10000)))
GMM_ASSERT1(false, "Error too large: " << error);
gmm::copy(gmm::conjugated(qaux), q2);
gmm::apply_house_left(gmm::transposed(m1), q2);
gmm::mult(gmm::conjugated(q2), gmm::transposed(dm1), dm1aux);
gmm::add(gmm::scaled(m1aux, T(-1)), dm1aux);
error = gmm::mat_euclidean_norm(dm1aux);
if (!(error <= prec * R(10000)))
GMM_ASSERT1(false, "Error too large: " << error);
}
//
// Test for implicit_qr_algorithm
//
m = gmm::mat_nrows(m2);
gmm::dense_matrix<T> cq(m, m), cr(m, m), ca(m, m);
std::vector<T> cv(m);
std::vector<std::complex<R> > eigc(m), cvc(m);
gmm::fill_random(ca);
std::complex<R> det1(gmm::lu_det(ca)), det2(1);
implicit_qr_algorithm(ca, eigc, cq);
for (size_type i = 0; i < m; ++i) det2 *= eigc[i];
if (gmm::abs(det1 - det2) > (gmm::abs(det1)+gmm::abs(det2))/R(50))
GMM_ASSERT1(false, "Error in QR or det. det lu: " << det1
<< " det qr: " << det2);
if (print_debug)
cout << "det lu = " << det1 << " det qr = " << det2 << endl;
if (m > 0) do {
gmm::fill_random(cq);
} while (gmm::abs(gmm::lu_det(cq)) < sqrt(prec)
|| gmm::condition_number(cq) > R(1000));
gmm::copy(cq, cr);
gmm::lu_inverse(cr);
gmm::fill_random(cv);
if (m > 0) cv[ 0] = real_or_complex( 0.0, 0.0, cv[0]);
if (m > 1) cv[ 1] = real_or_complex( 0.0, 0.0, cv[0]);
if (m > 2) cv[ 2] = real_or_complex( 0.01,-0.1, cv[0]);
if (m > 3) cv[ 3] = real_or_complex( 0.01, 0.1, cv[0]);
if (m > 4) cv[ 4] = real_or_complex( -2.0, 3.0, cv[0]);
if (m > 5) cv[ 5] = real_or_complex( -2.0, 3.0, cv[0]);
if (m > 6) cv[ 6] = real_or_complex( -50.0, 3.0, cv[0]);
if (m > 7) cv[ 7] = real_or_complex( 100.0, 1.0, cv[0]);
if (m > 8) cv[ 8] = real_or_complex( 300.0, 1.0, cv[0]);
if (m > 9) cv[ 9] = real_or_complex( 500.0, 1.0, cv[0]);
if (m > 10) cv[10] = real_or_complex( 1000.0, 1.0, cv[0]);
if (m > 11) cv[11] = real_or_complex( 4000.0, 1.0, cv[0]);
if (m > 12) cv[12] = real_or_complex( 5000.0, 1.0, cv[0]);
if (m > 13) cv[13] = real_or_complex( 10000.0, 1.0, cv[0]);
if (m > 14) cv[14] = real_or_complex( 80000.0, 1.0, cv[0]);
if (m > 15) cv[15] = real_or_complex(100000.0, 1.0, cv[0]);
gmm::clear(m2);
for (size_type l = 0; l < m; ++l) m2(l, l) = cv[l];
gmm::mult(cq, m2, ca);
gmm::mult(ca, cr, ca);
implicit_qr_algorithm(ca, eigc, cq);
gmm::copy(cv, cvc);
sort_eval(cvc);
sort_eval(eigc);
error = gmm::vect_dist2(cvc, eigc);
if (!(error <= sqrt(prec) * gmm::vect_norm2(cv) * R(10)))
GMM_ASSERT1(false, "Error in QR algorithm, error = " << error);
gmm::dense_matrix<T> aa(m, m), bb(m, m);
gmm::mult(gmm::conjugated(cq), ca, aa);
gmm::mult(aa, cq, bb);
for (size_type i = 0; i < m; ++i)
for (size_type j = (i == 0) ? 0 : i-1; j < m; ++j)
bb(i, j) = T(0);
error = gmm::mat_maxnorm(bb);
if (!(error <= sqrt(prec) * gmm::vect_norm2(cv) * R(10)))
GMM_ASSERT1(false, "Error in Schur vectors, error = "<< error);
//
// Test for symmetric_qr_algorithm
//
m = gmm::mat_nrows(m2);
std::vector<R> cvr(m), eigcr(m);
if (m > 0) do {
gmm::fill_random(cr);
} while (gmm::abs(gmm::lu_det(cr)) < sqrt(prec)
|| gmm::condition_number(cr) > R(1000));
gmm::qr_factor(cr, cq, ca);
gmm::fill_random(cvr);
gmm::copy(gmm::identity_matrix(), m2);
if (m > 0) cvr[ 0] = R( 0.0 );
if (m > 1) cvr[ 1] = R( 0.0 );
if (m > 2) cvr[ 2] = R( 0.01);
if (m > 3) cvr[ 3] = R( 0.01);
if (m > 4) cvr[ 4] = R( -2.0 );
if (m > 5) cvr[ 5] = R( -2.0 );
if (m > 6) cvr[ 6] = R( -50.0 );
if (m > 7) cvr[ 7] = R( 100.0 );
if (m > 8) cvr[ 8] = R( 300.0 );
if (m > 9) cvr[ 9] = R( 500.0 );
if (m > 10) cvr[10] = R( 1000.0 );
if (m > 11) cvr[11] = R( 4000.0 );
if (m > 12) cvr[12] = R( 5000.0 );
if (m > 13) cvr[13] = R( 10000.0 );
if (m > 14) cvr[14] = R( 80000.0 );
if (m > 15) cvr[15] = R(100000.0 );
gmm::clear(m2);
for (size_type l = 0; l < m; ++l) m2(l, l) = cvr[l];
gmm::mult(gmm::conjugated(cq), m2, ca);
gmm::mult(ca, cq, ca);
symmetric_qr_algorithm(ca, eigcr, cq);
for (size_type l = 0; l < m; ++l) {
std::vector<T> vy(m);
gmm::mult(ca, gmm::mat_col(cq, l),
gmm::scaled(gmm::mat_col(cq, l), -eigcr[l]), vy);
error = gmm::vect_norm2(vy);
if (!(error <= sqrt(prec) * gmm::vect_norm2(cvr) * R(10)))
GMM_ASSERT1(false, "Error too large: " << error);
}
sort_eval(cvr);
sort_eval(eigcr);
error = gmm::vect_dist2(cvr, eigcr);
if (!(error <= sqrt(prec) * gmm::vect_norm2(cvr) * R(10)))
GMM_ASSERT1(false, "Error in QR algorithm.");
if (nb_iter == 100) return true;
return false;
}
|
