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
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
|
// Copyright (C) 2007 Davis E. King (davis@dlib.net)
// License: Boost Software License See LICENSE.txt for the full license.
#undef DLIB_SVm_ABSTRACT_
#ifdef DLIB_SVm_ABSTRACT_
#include <cmath>
#include <limits>
#include <sstream>
#include "../matrix/matrix_abstract.h"
#include "../algs.h"
#include "../serialize.h"
#include "function_abstract.h"
#include "kernel_abstract.h"
#include "svm_nu_trainer_abstract.h"
namespace dlib
{
// ----------------------------------------------------------------------------------------
// ----------------------------------------------------------------------------------------
// ----------------------------------------------------------------------------------------
template <
typename T,
typename U
>
bool is_learning_problem (
const T& x,
const U& x_labels
);
/*!
requires
- T == a matrix or something convertible to a matrix via vector_to_matrix()
- U == a matrix or something convertible to a matrix via vector_to_matrix()
ensures
- returns true if all of the following are true and false otherwise:
- is_col_vector(x) == true
- is_col_vector(x_labels) == true
- x.size() == x_labels.size()
- x.size() > 0
!*/
template <
typename T,
typename U
>
bool is_binary_classification_problem (
const T& x,
const U& x_labels
);
/*!
requires
- T == a matrix or something convertible to a matrix via vector_to_matrix()
- U == a matrix or something convertible to a matrix via vector_to_matrix()
ensures
- returns true if all of the following are true and false otherwise:
- is_learning_problem(x, x_labels) == true
- x.size() > 1
- there exists at least one sample from both the +1 and -1 classes.
(i.e. all samples can't have the same label)
- for all valid i:
- x_labels(i) == -1 or +1
!*/
template <
typename sequence_type
>
bool is_sequence_labeling_problem (
const std::vector<sequence_type>& samples,
const std::vector<std::vector<unsigned long> >& labels
);
/*!
ensures
- returns true if all of the following are true and false otherwise:
- is_learning_problem(samples, labels) == true
- for all valid i:
- samples[i].size() == labels[i].size()
(i.e. The size of a label sequence need to match the size of
its corresponding sample sequence)
!*/
template <
typename lhs_type,
typename rhs_type
>
bool is_assignment_problem (
const std::vector<std::pair<std::vector<lhs_type>, std::vector<rhs_type> > >& samples,
const std::vector<std::vector<long> >& labels
);
/*!
ensures
- Note that an assignment problem is a task to associate each element of samples[i].first
to an element of samples[i].second, or to indicate that the element doesn't associate
with anything. Therefore, labels[i] should contain the association information for
samples[i].
- This function returns true if all of the following are true and false otherwise:
- is_learning_problem(samples, labels) == true
- for all valid i:
- samples[i].first.size() == labels[i].size()
- for all valid j:
-1 <= labels[i][j] < samples[i].second.size()
(A value of -1 indicates that samples[i].first[j] isn't associated with anything.
All other values indicate the associating element of samples[i].second)
- All elements of labels[i] which are not equal to -1 are unique. That is,
multiple elements of samples[i].first can't associate to the same element
in samples[i].second.
!*/
template <
typename lhs_type,
typename rhs_type
>
bool is_forced_assignment_problem (
const std::vector<std::pair<std::vector<lhs_type>, std::vector<rhs_type> > >& samples,
const std::vector<std::vector<long> >& labels
);
/*!
ensures
- A regular assignment problem is allowed to indicate that all elements of
samples[i].first don't associate to anything. However, a forced assignment
problem is required to always associate an element of samples[i].first to
something in samples[i].second if there is an element of samples[i].second
that hasn't already been associated to something.
- This function returns true if all of the following are true and false otherwise:
- is_assignment_problem(samples, labels) == true
- for all valid i:
- let N denote the number of elements in labels[i] that are not equal to -1.
- min(samples[i].first.size(), samples[i].second.size()) == N
!*/
// ----------------------------------------------------------------------------------------
// ----------------------------------------------------------------------------------------
// ----------------------------------------------------------------------------------------
template <
typename trainer_type,
typename sample_vector_type,
typename label_vector_type
>
const probabilistic_function<typename trainer_type::trained_function_type>
train_probabilistic_decision_function (
const trainer_type& trainer,
const sample_vector_type& x,
const label_vector_type& y,
const long folds
);
/*!
requires
- 1 < folds <= x.size()
- is_binary_classification_problem(x,y) == true
- x and y must be std::vector objects or types with a compatible interface.
- trainer_type == some kind of batch trainer object (e.g. svm_nu_trainer)
ensures
- trains a classifier given the training samples in x and labels in y.
- returns a probabilistic_decision_function that represents the trained classifier.
- The parameters of the probability model are estimated by performing k-fold
cross validation.
- The number of folds used is given by the folds argument.
throws
- any exceptions thrown by trainer.train()
- std::bad_alloc
!*/
// ----------------------------------------------------------------------------------------
template <
typename trainer_type
>
trainer_adapter_probabilistic<trainer_type> probabilistic (
const trainer_type& trainer,
const long folds
);
/*!
requires
- 1 < folds <= x.size()
- trainer_type == some kind of batch trainer object (e.g. svm_nu_trainer)
ensures
- returns a trainer adapter TA such that calling TA.train(samples, labels)
returns the same object as calling train_probabilistic_decision_function(trainer,samples,labels,folds).
!*/
// ----------------------------------------------------------------------------------------
// ----------------------------------------------------------------------------------------
// Miscellaneous functions
// ----------------------------------------------------------------------------------------
// ----------------------------------------------------------------------------------------
template <
typename trainer_type,
typename in_sample_vector_type,
typename in_scalar_vector_type
>
const matrix<double,1,2> cross_validate_trainer (
const trainer_type& trainer,
const in_sample_vector_type& x,
const in_scalar_vector_type& y,
const long folds
);
/*!
requires
- is_binary_classification_problem(x,y) == true
- 1 < folds <= x.nr()
- trainer_type == some kind of binary classification trainer object (e.g. svm_nu_trainer)
ensures
- performs k-fold cross validation by using the given trainer to solve the
given binary classification problem for the given number of folds.
Each fold is tested using the output of the trainer and the average
classification accuracy from all folds is returned.
- The average accuracy is computed by running test_binary_decision_function()
on each fold and its output is averaged and returned.
- The number of folds used is given by the folds argument.
throws
- any exceptions thrown by trainer.train()
- std::bad_alloc
!*/
// ----------------------------------------------------------------------------------------
template <
typename dec_funct_type,
typename in_sample_vector_type,
typename in_scalar_vector_type
>
const matrix<double,1,2> test_binary_decision_function (
const dec_funct_type& dec_funct,
const in_sample_vector_type& x_test,
const in_scalar_vector_type& y_test
);
/*!
requires
- is_binary_classification_problem(x_test,y_test) == true
- dec_funct_type == some kind of decision function object (e.g. decision_function)
ensures
- Tests the given decision function by calling it on the x_test and y_test samples.
The output of dec_funct is interpreted as a prediction for the +1 class
if its output is >= 0 and as a prediction for the -1 class otherwise.
- The test accuracy is returned in a row vector, let us call it R. Both
quantities in R are numbers between 0 and 1 which represent the fraction
of examples correctly classified. R(0) is the fraction of +1 examples
correctly classified and R(1) is the fraction of -1 examples correctly
classified.
throws
- std::bad_alloc
!*/
// ----------------------------------------------------------------------------------------
// ----------------------------------------------------------------------------------------
template <
typename T,
typename U
>
void randomize_samples (
T& samples,
U& labels
);
/*!
requires
- T == a matrix object or an object compatible with std::vector that contains
a swappable type.
- U == a matrix object or an object compatible with std::vector that contains
a swappable type.
- if samples or labels are matrix objects then is_vector(samples) == true and
is_vector(labels) == true
- samples.size() == labels.size()
ensures
- randomizes the order of the samples and labels but preserves
the pairing between each sample and its label
- A default initialized random number generator is used to perform the randomizing.
Note that this means that each call this this function does the same thing.
That is, the random number generator always uses the same seed.
- for all valid i:
- let r == the random index samples(i) was moved to. then:
- #labels(r) == labels(i)
!*/
// ----------------------------------------------------------------------------------------
template <
typename T,
typename U,
typename rand_type
>
void randomize_samples (
T& samples,
U& labels,
rand_type& rnd
);
/*!
requires
- T == a matrix object or an object compatible with std::vector that contains
a swappable type.
- U == a matrix object or an object compatible with std::vector that contains
a swappable type.
- if samples or labels are matrix objects then is_vector(samples) == true and
is_vector(labels) == true
- samples.size() == labels.size()
- rand_type == a type that implements the dlib/rand/rand_kernel_abstract.h interface
ensures
- randomizes the order of the samples and labels but preserves
the pairing between each sample and its label
- the given rnd random number generator object is used to do the randomizing
- for all valid i:
- let r == the random index samples(i) was moved to. then:
- #labels(r) == labels(i)
!*/
// ----------------------------------------------------------------------------------------
template <
typename T
>
void randomize_samples (
T& samples
);
/*!
requires
- T == a matrix object or an object compatible with std::vector that contains
a swappable type.
- if samples is a matrix then is_vector(samples) == true
ensures
- randomizes the order of the elements inside samples
- A default initialized random number generator is used to perform the randomizing.
Note that this means that each call this this function does the same thing.
That is, the random number generator always uses the same seed.
!*/
// ----------------------------------------------------------------------------------------
template <
typename T,
typename rand_type
>
void randomize_samples (
T& samples,
rand_type& rnd
);
/*!
requires
- T == a matrix object or an object compatible with std::vector that contains
a swappable type.
- if samples is a matrix then is_vector(samples) == true
ensures
- randomizes the order of the elements inside samples
- the given rnd random number generator object is used to do the randomizing
!*/
// ----------------------------------------------------------------------------------------
template <
typename T,
typename U,
typename V
>
void randomize_samples (
T& samples,
U& labels,
V& auxiliary
);
/*!
requires
- T == a matrix object or an object compatible with std::vector that contains
a swappable type.
- U == a matrix object or an object compatible with std::vector that contains
a swappable type.
- V == a matrix object or an object compatible with std::vector that contains
a swappable type.
- if (samples, labels, or auxiliary are matrix objects) then
- is_vector(samples) == true
- is_vector(labels) == true
- is_vector(auxiliary) == true
- samples.size() == labels.size() == auxiliary.size()
- rand_type == a type that implements the dlib/rand/rand_kernel_abstract.h interface
ensures
- randomizes the order of the samples, labels, and auxiliary but preserves the
pairing between each sample, its label, and its auxiliary value.
- A default initialized random number generator is used to perform the
randomizing. Note that this means that each call this this function does the
same thing. That is, the random number generator always uses the same seed.
- for all valid i:
- let r == the random index samples(i) was moved to. then:
- #labels(r) == labels(i)
- #auxiliary(r) == auxiliary(i)
!*/
// ----------------------------------------------------------------------------------------
template <
typename T,
typename U,
typename V,
typename rand_type
>
void randomize_samples (
T& samples,
U& labels,
V& auxiliary,
rand_type& rnd
);
/*!
requires
- T == a matrix object or an object compatible with std::vector that contains
a swappable type.
- U == a matrix object or an object compatible with std::vector that contains
a swappable type.
- V == a matrix object or an object compatible with std::vector that contains
a swappable type.
- if (samples, labels, or auxiliary are matrix objects) then
- is_vector(samples) == true
- is_vector(labels) == true
- is_vector(auxiliary) == true
- samples.size() == labels.size() == auxiliary.size()
- rand_type == a type that implements the dlib/rand/rand_kernel_abstract.h interface
ensures
- randomizes the order of the samples, labels, and auxiliary but preserves the
pairing between each sample, its label, and its auxiliary value.
- the given rnd random number generator object is used to do the randomizing
- for all valid i:
- let r == the random index samples(i) was moved to. then:
- #labels(r) == labels(i)
- #auxiliary(r) == auxiliary(i)
!*/
// ----------------------------------------------------------------------------------------
// ----------------------------------------------------------------------------------------
}
#endif // DLIB_SVm_ABSTRACT_
|
