File: cross_validate_assignment_trainer.h

package info (click to toggle)
mldemos 0.5.1-3
  • links: PTS, VCS
  • area: main
  • in suites: jessie, jessie-kfreebsd
  • size: 32,224 kB
  • ctags: 46,525
  • sloc: cpp: 306,887; ansic: 167,718; ml: 126; sh: 109; makefile: 2
file content (181 lines) | stat: -rw-r--r-- 6,548 bytes parent folder | download | duplicates (4)
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
// Copyright (C) 2011  Davis E. King (davis@dlib.net)
// License: Boost Software License   See LICENSE.txt for the full license.
#ifndef DLIB_CROSS_VALIDATE_ASSiGNEMNT_TRAINER_H__
#define DLIB_CROSS_VALIDATE_ASSiGNEMNT_TRAINER_H__

#include "cross_validate_assignment_trainer_abstract.h"
#include <vector>
#include "../matrix.h"
#include "svm.h"


namespace dlib
{

// ----------------------------------------------------------------------------------------

    template <
        typename assignment_function
        >
    double test_assignment_function (
        const assignment_function& assigner,
        const std::vector<typename assignment_function::sample_type>& samples,
        const std::vector<typename assignment_function::label_type>& labels
    )
    {
        // make sure requires clause is not broken
#ifdef ENABLE_ASSERTS
        if (assigner.forces_assignment())
        {
            DLIB_ASSERT(is_forced_assignment_problem(samples, labels),
                "\t double test_assignment_function()"
                << "\n\t invalid inputs were given to this function"
                << "\n\t is_forced_assignment_problem(samples,labels): " << is_forced_assignment_problem(samples,labels)
                << "\n\t is_assignment_problem(samples,labels):        " << is_assignment_problem(samples,labels)
                << "\n\t is_learning_problem(samples,labels):          " << is_learning_problem(samples,labels)
                );
        }
        else
        {
            DLIB_ASSERT(is_assignment_problem(samples, labels),
                "\t double test_assignment_function()"
                << "\n\t invalid inputs were given to this function"
                << "\n\t is_assignment_problem(samples,labels): " << is_assignment_problem(samples,labels)
                << "\n\t is_learning_problem(samples,labels):   " << is_learning_problem(samples,labels)
                );
        }
#endif
        double total_right = 0;
        double total = 0;
        for (unsigned long i = 0; i < samples.size(); ++i)
        {
            const std::vector<long>& out = assigner(samples[i]);
            for (unsigned long j = 0; j < out.size(); ++j)
            {
                if (out[j] == labels[i][j])
                    ++total_right;

                ++total;
            }
        }

        if (total != 0)
            return total_right/total;
        else
            return 1;
    }

// ----------------------------------------------------------------------------------------

    template <
        typename trainer_type
        >
    double cross_validate_assignment_trainer (
        const trainer_type& trainer,
        const std::vector<typename trainer_type::sample_type>& samples,
        const std::vector<typename trainer_type::label_type>& labels,
        const long folds
    )
    {
        // make sure requires clause is not broken
#ifdef ENABLE_ASSERTS
        if (trainer.forces_assignment())
        {
            DLIB_ASSERT(is_forced_assignment_problem(samples, labels) &&
                        1 < folds && folds <= static_cast<long>(samples.size()),
                "\t double cross_validate_assignment_trainer()"
                << "\n\t invalid inputs were given to this function"
                << "\n\t samples.size(): " << samples.size() 
                << "\n\t folds:  " << folds 
                << "\n\t is_forced_assignment_problem(samples,labels): " << is_forced_assignment_problem(samples,labels)
                << "\n\t is_assignment_problem(samples,labels):        " << is_assignment_problem(samples,labels)
                << "\n\t is_learning_problem(samples,labels):          " << is_learning_problem(samples,labels)
                );
        }
        else
        {
            DLIB_ASSERT(is_assignment_problem(samples, labels) &&
                        1 < folds && folds <= static_cast<long>(samples.size()),
                "\t double cross_validate_assignment_trainer()"
                << "\n\t invalid inputs were given to this function"
                << "\n\t samples.size(): " << samples.size() 
                << "\n\t folds:  " << folds 
                << "\n\t is_assignment_problem(samples,labels): " << is_assignment_problem(samples,labels)
                << "\n\t is_learning_problem(samples,labels):   " << is_learning_problem(samples,labels)
                );
        }
#endif



        typedef typename trainer_type::sample_type sample_type;
        typedef typename trainer_type::label_type label_type;

        const long num_in_test  = samples.size()/folds;
        const long num_in_train = samples.size() - num_in_test;


        std::vector<sample_type> samples_test, samples_train;
        std::vector<label_type> labels_test, labels_train;


        long next_test_idx = 0;
        double total_right = 0;
        double total = 0;


        for (long i = 0; i < folds; ++i)
        {
            samples_test.clear();
            labels_test.clear();
            samples_train.clear();
            labels_train.clear();

            // load up the test samples
            for (long cnt = 0; cnt < num_in_test; ++cnt)
            {
                samples_test.push_back(samples[next_test_idx]);
                labels_test.push_back(labels[next_test_idx]);
                next_test_idx = (next_test_idx + 1)%samples.size();
            }

            // load up the training samples
            long next = next_test_idx;
            for (long cnt = 0; cnt < num_in_train; ++cnt)
            {
                samples_train.push_back(samples[next]);
                labels_train.push_back(labels[next]);
                next = (next + 1)%samples.size();
            }


            const typename trainer_type::trained_function_type& df = trainer.train(samples_train,labels_train);

            // check how good df is on the test data
            for (unsigned long i = 0; i < samples_test.size(); ++i)
            {
                const std::vector<long>& out = df(samples_test[i]);
                for (unsigned long j = 0; j < out.size(); ++j)
                {
                    if (out[j] == labels_test[i][j])
                        ++total_right;

                    ++total;
                }
            }

        } // for (long i = 0; i < folds; ++i)

        if (total != 0)
            return total_right/total;
        else
            return 1;

    }

// ----------------------------------------------------------------------------------------

}

#endif // DLIB_CROSS_VALIDATE_ASSiGNEMNT_TRAINER_H__