forked from johannesgerer/jburkardt-f
-
Notifications
You must be signed in to change notification settings - Fork 1
/
qr_solve.html
347 lines (303 loc) · 11 KB
/
qr_solve.html
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
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
<html>
<head>
<title>
QR_SOLVE - Least Squares Solution of a Linear System A*x=b
</title>
</head>
<body bgcolor="#EEEEEE" link="#CC0000" alink="#FF3300" vlink="#000055">
<h1 align = "center">
QR_SOLVE <br> Least Squares Solution of a Linear System A*x=b
</h1>
<hr>
<p>
<b>QR_SOLVE</b>
is a FORTRAN90 library which
computes a least squares solution of a linear system of the form A*x=b.
</p>
<p>
There are many possible cases that can arise with the matrix A.
Formally, we distinguish the cases M < N, M = N, and M > N, and
we expect trouble whenever M is not equal to N. Trouble may also
arise when M = N but the matrix is singular. Even if the matrix is,
mathematically speaking, non-singular, it may be so close to singularity
that an accurate solution is difficult to achieve.
</p>
<p>
When M > N, we are placing more conditions than we have degrees of freedom,
so we suppose that such a linear system cannot be solved. However, it is
possible that the extra conditions are illusory, being constructed from
linear combinations of a fundamental set of N conditions. Thus, a system
that we typically call "overdetermined" can have a solution in the ordinary
sense, that satisfies all the conditions, as long as the right hand side is
"consistent". Another way of saying this is that the system is solvable
if the right hand side lies in the column space of A...although that simply
says that it is a linear combination of the columns of A, which just says
A*x=b.
</p>
<p>
If A does not have full column rank, however, then even if the right hand
side lies in the column space of A, there will be more than one linear
combination of columns that produce b. Thus, the equations are consistent,
the system is solvable, but not uniquely so.
</p>
<p>
If M < N, then we are placing fewer conditions than we have degrees of
freedom. As long as the right hand side lies in the column space of A,
we can guarantee that there will be multiple solutions.
</p>
<p>
Thus, the question of a "solution" to the problem A*x=b is complicated
enough that it seems to defy a common algorithmic approach. Nonetheless,
there are some sensible, robust procedures for producing an answer that
corresponds to the classical solution, or solves the overdetermined problem
when the right hand side is consistent. This is the linear least squares
solution, which finds a vector x which minimizes the Euclidean norm of
the residual: ||r|| = ||A*x-b||. This solution is produced by computing
the QR factorization of the matrix A
</p>
<p>
When there are multiple solutions to the problem, the QR approach used here
produces a solution. A more satisfactory approach, using the pseudoinverse,
will produce a solution x which satisfies the additional constraint that
it has the minimum norm of the entire family of solutions. That pseudoinverse
approach is not implemented in this library. The singular value decomposition (SVD)
can also produce this minimal solution.
</p>
<p>
For comparison, a solver that applies the normal equations is included.
This approach requires M >= N, and that A have full column rank. It constructs
and solves the NxN system A'*A*x=A'*b. This system has a condition number that
is the square of the original system, so it also suffers from a significant
loss in accuracy.
</p>
<p>
We also include an SVD solver, which uses the pseudoinverse approach.
First compute A = U * S * V', where U and V are orthogonal, and S is
MxN diagonal, then to solve A*x=b write x = V * S^ * U' * b, where
S^ is the matrix formed by transposing S and then replacing each nonzero
diagonal element s by 1/s. (However, very small elements should probably
be zeroed instead of inverted.) This procedure will also produce a vector
x which minimizes the Euclidean norm. However, it has one feature that the
QR solver does not: in cases where the solution x is not unique (because
A does not have full column rank) the SVD solver returns the unique vector
x of minimum Euclidean norm.
</p>
<p>
<b>QR_SOLVE</b> needs the R8LIB library. The test program also needs
the TEST_LS library.
</p>
<h3 align = "center">
Licensing:
</h3>
<p>
The computer code and data files described and made available on this
web page are distributed under
<a href = "../../txt/gnu_lgpl.txt">the GNU LGPL license.</a>
</p>
<h3 align = "center">
Languages:
</h3>
<p>
<b>QR_SOLVE</b> is available in
<a href = "../../c_src/qr_solve/qr_solve.html">a C version</a> and
<a href = "../../cpp_src/qr_solve/qr_solve.html">a C++ version</a> and
<a href = "../../f77_src/qr_solve/qr_solve.html">a FORTRAN77 version</a> and
<a href = "../../f_src/qr_solve/qr_solve.html">a FORTRAN90 version</a>.
</p>
<h3 align = "center">
Related Data and Programs:
</h3>
<p>
<a href = "../../f_src/bvls/bvls.html">
BVLS</a>,
a FORTRAN90 library which
applies least squares methods to solve a linear system for which
lower and upper constraints may have been placed on every variable,
by Charles Lawson and Richard Hanson.
</p>
<p>
<a href = "../../f_src/dqed/dqed.html">
DQED</a>,
a FORTRAN90 library which
solves constrained least squares problems,
by Richard Hanson and Fred Krogh.
</p>
<p>
<a href = "../../f_src/lapack_examples/lapack_examples.html">
LAPACK_EXAMPLES</a>,
a FORTRAN90 program which
demonstrates the use of the LAPACK linear algebra library.
</p>
<p>
<a href = "../../f_src/lawson/lawson.html">
LAWSON</a>,
a FORTRAN90 library which
contains routines for solving least squares problems and singular value
decompositions (SVD),
by Charles Lawson, Richard Hanson.
</p>
<p>
<a href = "../../f_src/linpack_d/linpack_d.html">
LINPACK_D</a>,
a FORTRAN90 library which
solves linear systems using double precision real arithmetic;
</p>
<p>
<a href = "../../f_src/llsq/llsq.html">
LLSQ</a>,
a FORTRAN90 library which
solves the simple linear least squares problem of finding the formula
of a straight line y=a*x+b which minimizes the root-mean-square error
to a set of N data points.
</p>
<p>
<a href = "../../f_src/minpack/minpack.html">
MINPACK</a>,
a FORTRAN90 library which
solves systems of nonlinear equations, or the least squares minimization of the
residual of a set of linear or nonlinear equations.
</p>
<p>
<a href = "../../f_src/nms/nms.html">
NMS</a>,
a FORTRAN90 library which
includes a wide variety of numerical software, including
solvers for linear systems of equations, interpolation of data,
numerical quadrature, linear least squares data fitting,
the solution of nonlinear equations, ordinary differential equations,
optimization and nonlinear least squares, simulation and random numbers,
trigonometric approximation and Fast Fourier Transforms (FFT).
</p>
<p>
<a href = "../../f_src/r8lib/r8lib.html">
R8LIB</a>,
a FORTRAN90 library which
contains many utility routines using double precision real (R8) arithmetic.
</p>
<p>
<a href = "../../f_src/test_ls/test_ls.html">
TEST_LS</a>,
a FORTRAN90 library which
implements linear least squares test problems of the form A*x=b.
</p>
<h3 align = "center">
Reference:
</h3>
<p>
<ol>
<li>
Jack Dongarra, Jim Bunch, Cleve Moler, Pete Stewart,<br>
LINPACK User's Guide,<br>
SIAM, 1979,<br>
ISBN13: 978-0-898711-72-1,<br>
LC: QA214.L56.
</li>
<li>
David Kahaner, Cleve Moler, Steven Nash,<br>
Numerical Methods and Software,<br>
Prentice Hall, 1989,<br>
ISBN: 0-13-627258-4,<br>
LC: TA345.K34.
</li>
</ol>
</p>
<h3 align = "center">
Source Code:
</h3>
<p>
<ul>
<li>
<a href = "qr_solve.f90">qr_solve.f90</a>, the source code.
</li>
<li>
<a href = "qr_solve.sh">qr_solve.sh</a>,
BASH commands to compile the source code.
</li>
</ul>
</p>
<h3 align = "center">
Examples and Tests:
</h3>
<p>
<ul>
<li>
<a href = "qr_solve_prb.f90">qr_solve_prb.f90</a>,
a sample calling program.
</li>
<li>
<a href = "qr_solve_prb.sh">qr_solve_prb.sh</a>,
BASH commands to compile and run the sample program.
</li>
<li>
<a href = "qr_solve_prb_output.txt">qr_solve_prb_output.txt</a>,
the output file.
</li>
</ul>
</p>
<h3 align = "center">
List of Routines:
</h3>
<p>
<ul>
<li>
<b>DAXPY</b> computes constant times a vector plus a vector.
</li>
<li>
<b>DDOT</b> forms the dot product of two vectors.
</li>
<li>
<b>DNRM2</b> returns the euclidean norm of a vector.
</li>
<li>
<b>DQRANK</b> computes the QR factorization of a rectangular matrix.
</li>
<li>
<b>DQRDC</b> computes the QR factorization of a real rectangular matrix.
</li>
<li>
<b>DQRLS</b> factors and solves a linear system in the least squares sense.
</li>
<li>
<b>DQRLSS</b> solves a linear system in a least squares sense.
</li>
<li>
<b>DQRSL</b> computes transformations, projections, and least squares solutions.
</li>
<li>
<b>DROT</b> applies a plane rotation.
</li>
<li>
<b>DROTG</b> constructs a Givens plane rotation.
</li>
<li>
<b>DSCAL</b> scales a vector by a constant.
</li>
<li>
<b>DSVDC</b> computes the singular value decomposition of a real rectangular matrix.
</li>
<li>
<b>DSWAP</b> interchanges two vectors.
</li>
<li>
<b>NORMAL_SOLVE</b> solves a linear system using the normal equations.
</li>
<li>
<b>QR_SOLVE</b> solves a linear system in the least squares sense.
</li>
<li>
<b>SVD_SOLVE</b> solves a linear system in the least squares sense.
</li>
</ul>
</p>
<p>
You can go up one level to <a href = "../f_src.html">
the FORTRAN90 source codes</a>.
</p>
<hr>
<i>
Last revised on 04 October 2012.
</i>
<!-- John Burkardt -->
</body>
<!-- Initial HTML skeleton created by HTMLINDEX. -->
</html>