-
Notifications
You must be signed in to change notification settings - Fork 26
/
pca.cc
345 lines (323 loc) · 13.4 KB
/
pca.cc
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
// ------------------------------------------------------------------------ //
// Copyright 2021 SPTK Working Group //
// //
// Licensed under the Apache License, Version 2.0 (the "License"); //
// you may not use this file except in compliance with the License. //
// You may obtain a copy of the License at //
// //
// http://www.apache.org/licenses/LICENSE-2.0 //
// //
// Unless required by applicable law or agreed to in writing, software //
// distributed under the License is distributed on an "AS IS" BASIS, //
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. //
// See the License for the specific language governing permissions and //
// limitations under the License. //
// ------------------------------------------------------------------------ //
#include <algorithm> // std::transform
#include <fstream> // std::ifstream, std::ofstream
#include <iomanip> // std::setw
#include <iostream> // std::cerr, std::cin, std::cout, std::endl, etc.
#include <numeric> // std::accumulate
#include <sstream> // std::ostringstream
#include <vector> // std::vector
#include "GETOPT/ya_getopt.h"
#include "SPTK/math/principal_component_analysis.h"
#include "SPTK/utils/sptk_utils.h"
namespace {
const int kDefaultVectorLength(25);
const int kDefaultNumPrincipalComponent(2);
const int kDefaultNumIteration(10000);
const double kDefaultConvergenceThreshold(1e-6);
const sptk::PrincipalComponentAnalysis::CovarianceType kDefaultCovarianceType(
sptk::PrincipalComponentAnalysis::CovarianceType::kSampleCovariance);
void PrintUsage(std::ostream* stream) {
// clang-format off
*stream << std::endl;
*stream << " pca - principal component analysis" << std::endl;
*stream << std::endl;
*stream << " usage:" << std::endl;
*stream << " pca [ options ] [ infile ] > stdout" << std::endl;
*stream << " options:" << std::endl;
*stream << " -l l : length of vector ( int)[" << std::setw(5) << std::right << kDefaultVectorLength << "][ 1 <= l <= ]" << std::endl; // NOLINT
*stream << " -m m : order of vector ( int)[" << std::setw(5) << std::right << "l-1" << "][ 0 <= m <= ]" << std::endl; // NOLINT
*stream << " -n n : number of principal components ( int)[" << std::setw(5) << std::right << kDefaultNumPrincipalComponent << "][ 1 <= n <= l ]" << std::endl; // NOLINT
*stream << " -i i : maximum number of iterations ( int)[" << std::setw(5) << std::right << kDefaultNumIteration << "][ 1 <= i <= ]" << std::endl; // NOLINT
*stream << " -d d : convergence threshold (double)[" << std::setw(5) << std::right << kDefaultConvergenceThreshold << "][ 0.0 <= d <= ]" << std::endl; // NOLINT
*stream << " -u u : covariance type ( int)[" << std::setw(5) << std::right << kDefaultCovarianceType << "][ 0 <= u <= 2 ]" << std::endl; // NOLINT
*stream << " 0 (sample covariance)" << std::endl;
*stream << " 1 (unbiased covariance)" << std::endl;
*stream << " 2 (correlation)" << std::endl;
*stream << " -v v : output filename of double type (string)[" << std::setw(5) << std::right << "N/A" << "]" << std::endl; // NOLINT
*stream << " eigenvalues and proportions" << std::endl;
*stream << " -h : print this message" << std::endl;
*stream << " infile:" << std::endl;
*stream << " vector sequence (double)[stdin]" << std::endl; // NOLINT
*stream << " stdout:" << std::endl;
*stream << " mean vector and eigenvectors (double)" << std::endl; // NOLINT
*stream << std::endl;
*stream << " SPTK: version " << sptk::kVersion << std::endl;
*stream << std::endl;
// clang-format on
}
} // namespace
/**
* @a pca [ @e option ] [ @e infile ]
*
* - @b -l @e int
* - length of vector @f$(1 \le L)@f$
* - @b -m @e int
* - order of vector @f$(0 \le M)@f$
* - @b -n @e int
* - number of principal components @f$(1 \le N \le L)@f$
* - @b -i @e int
* - number of iterations @f$(1 \le I)@f$
* - @b -d @e double
* - convergence threshold @f$(0 \le \epsilon)@f$
* - @b -u @e int
* - covariance type
* @arg @c 0 sample covariance
* @arg @c 1 unbiased covariance
* @arg @c 2 correlation
* - @b -v @e str
* - double-type eigenvalues and proportions
* - @b infile @e str
* - double-type vector sequence
* - @b stdout
* - double-type mean vector and eigenvectors
*
* In the below example, principal component analysis is applied to the
* three-dimensional training vectors contained in @c data.d.
* The eigenvectors and the eigenvalues are written to @c eigvec.dat and
* @c eigval.dat, repectively.
*
* @code{.sh}
* pca -l 3 -n 2 -v eigval.dat < data.d > eigvec.dat
* @endcode
*
* The eigenvalues are sorted in descending order.
*
* @param[in] argc Number of arguments.
* @param[in] argv Argument vector.
* @return 0 on success, 1 on failure.
*/
int main(int argc, char* argv[]) {
int vector_length(kDefaultVectorLength);
int num_principal_component(kDefaultNumPrincipalComponent);
int num_iteration(kDefaultNumIteration);
double convergence_threshold(kDefaultConvergenceThreshold);
sptk::PrincipalComponentAnalysis::CovarianceType covariance_type(
kDefaultCovarianceType);
const char* eigenvalues_file(NULL);
for (;;) {
const int option_char(
getopt_long(argc, argv, "l:m:n:i:d:u:v:h", NULL, NULL));
if (-1 == option_char) break;
switch (option_char) {
case 'l': {
if (!sptk::ConvertStringToInteger(optarg, &vector_length) ||
vector_length <= 0) {
std::ostringstream error_message;
error_message
<< "The argument for the -l option must be a positive integer";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
break;
}
case 'm': {
if (!sptk::ConvertStringToInteger(optarg, &vector_length) ||
vector_length < 0) {
std::ostringstream error_message;
error_message << "The argument for the -m option must be a "
<< "non-negative integer";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
++vector_length;
break;
}
case 'n': {
if (!sptk::ConvertStringToInteger(optarg, &num_principal_component) ||
num_principal_component <= 0) {
std::ostringstream error_message;
error_message
<< "The argument for the -n option must be a positive integer";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
break;
}
case 'i': {
if (!sptk::ConvertStringToInteger(optarg, &num_iteration) ||
num_iteration <= 0) {
std::ostringstream error_message;
error_message
<< "The argument for the -i option must be a positive integer";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
break;
}
case 'd': {
if (!sptk::ConvertStringToDouble(optarg, &convergence_threshold) ||
convergence_threshold < 0.0) {
std::ostringstream error_message;
error_message
<< "The argument for the -d option must be a non-negative number";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
break;
}
case 'u': {
const int min(0);
const int max(
static_cast<int>(sptk::PrincipalComponentAnalysis::CovarianceType::
kNumCovarianceTypes) -
1);
int tmp;
if (!sptk::ConvertStringToInteger(optarg, &tmp) ||
!sptk::IsInRange(tmp, min, max)) {
std::ostringstream error_message;
error_message << "The argument for the -u option must be an integer "
<< "in the range of " << min << " to " << max;
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
covariance_type =
static_cast<sptk::PrincipalComponentAnalysis::CovarianceType>(tmp);
break;
}
case 'v': {
eigenvalues_file = optarg;
break;
}
case 'h': {
PrintUsage(&std::cout);
return 0;
}
default: {
PrintUsage(&std::cerr);
return 1;
}
}
}
if (vector_length < num_principal_component) {
std::ostringstream error_message;
error_message << "Number of principal components must be equal to or "
<< "less than length of input vector";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
const int num_input_files(argc - optind);
if (1 < num_input_files) {
std::ostringstream error_message;
error_message << "Too many input files";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
const char* input_file(0 == num_input_files ? NULL : argv[optind]);
if (!sptk::SetBinaryMode()) {
std::ostringstream error_message;
error_message << "Cannot set translation mode";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
// Open stream for reading inputs.
std::ifstream ifs;
if (NULL != input_file) {
ifs.open(input_file, std::ios::in | std::ios::binary);
if (ifs.fail()) {
std::ostringstream error_message;
error_message << "Cannot open file " << input_file;
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
}
std::istream& input_stream(ifs.is_open() ? ifs : std::cin);
// Open stream for writing eigenvalues.
std::ofstream ofs;
if (NULL != eigenvalues_file) {
ofs.open(eigenvalues_file, std::ios::out | std::ios::binary);
if (ofs.fail()) {
std::ostringstream error_message;
error_message << "Cannot open file " << eigenvalues_file;
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
}
std::ostream& output_stream(ofs);
sptk::PrincipalComponentAnalysis principal_component_analysis(
vector_length - 1, num_iteration, convergence_threshold, covariance_type);
sptk::PrincipalComponentAnalysis::Buffer buffer;
if (!principal_component_analysis.IsValid()) {
std::ostringstream error_message;
error_message << "Failed to initialize PrincipalComponentAnalysis";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
// Read input data.
std::vector<std::vector<double> > input_vectors;
{
std::vector<double> tmp(vector_length);
while (sptk::ReadStream(false, 0, 0, vector_length, &tmp, &input_stream,
NULL)) {
input_vectors.push_back(tmp);
}
}
if (input_vectors.empty()) return 0;
std::vector<double> mean_vector(vector_length);
std::vector<double> eigenvalues(vector_length);
sptk::Matrix eigenvector_matrix(vector_length, vector_length);
if (!principal_component_analysis.Run(input_vectors, &mean_vector,
&eigenvalues, &eigenvector_matrix,
&buffer)) {
std::ostringstream error_message;
error_message << "Failed to perform principal component analysis";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
if (!sptk::WriteStream(0, vector_length, mean_vector, &std::cout, NULL)) {
std::ostringstream error_message;
error_message << "Failed to write mean vector";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
sptk::Matrix eigenvector_submatrix;
if (!eigenvector_matrix.GetSubmatrix(0, num_principal_component, 0,
vector_length, &eigenvector_submatrix)) {
std::ostringstream error_message;
error_message << "Failed to get eigenvectors";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
if (!sptk::WriteStream(eigenvector_submatrix, &std::cout)) {
std::ostringstream error_message;
error_message << "Failed to write eigenvectors";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
if (NULL != eigenvalues_file) {
if (!sptk::WriteStream(0, num_principal_component, eigenvalues,
&output_stream, NULL)) {
std::ostringstream error_message;
error_message << "Failed to write eigenvalues";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
const double norm(
1.0 / std::accumulate(eigenvalues.begin(), eigenvalues.end(), 0.0));
std::vector<double> proportions(num_principal_component);
std::transform(eigenvalues.begin(),
eigenvalues.begin() + num_principal_component,
proportions.begin(), [norm](double l) { return l * norm; });
if (!sptk::WriteStream(0, num_principal_component, proportions,
&output_stream, NULL)) {
std::ostringstream error_message;
error_message << "Failed to write proportions";
sptk::PrintErrorMessage("pca", error_message);
return 1;
}
}
return 0;
}