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
|
// Copyright (C) 2002 Samy Bengio (bengio@idiap.ch)
//
//
// This file is part of Torch. Release II.
// [The Ultimate Machine Learning Library]
//
// Torch is free software; you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation; either version 2 of the License, or
// (at your option) any later version.
//
// Torch 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 General Public License for more details.
//
// You should have received a copy of the GNU General Public License
// along with Torch; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
#include "Bagging.h"
#include "random.h"
namespace Torch {
Bagging::Bagging(WeightedSumMachine* w_machine_, DataSet* data_) : Trainer(w_machine_,data_)
{
w_machine = w_machine_;
int n_examples = data->n_examples;
n_trainers = w_machine->n_trainers;
is_selected_examples = (int *)xalloc(n_examples*sizeof(int));
n_unselected_examples = (int *)xalloc(n_trainers*sizeof(int));
unselected_examples = (int **)xalloc(n_trainers*sizeof(int*));
selected_examples = (int **)xalloc(n_trainers*sizeof(int*));
for (int i=0;i<n_trainers;i++) {
unselected_examples[i] = (int *)xalloc(n_examples*sizeof(int));
selected_examples[i] = (int *)xalloc(n_examples*sizeof(int));
}
for(int i = 0; i < n_trainers; i++)
w_machine->weights[i] = 1./((real)n_trainers);
}
void Bagging::bootstrapData(int* selected, int* is_selected)
{
int n = data->n_examples;
for (int j=0;j<n;j++) {
selected[j] = (int)floor(bounded_uniform(0,n));
is_selected[selected[j]] = 1;
}
}
void Bagging::train(List* measurers)
{
message("Bagging: training");
w_machine->n_trainers_trained = 0;
int n = data->n_examples;
for (int i=0;i<n_trainers;i++) {
// initialization
for (int j=0;j<n;j++) {
is_selected_examples[j]=0;
}
// select a bootstrap
bootstrapData(selected_examples[i],is_selected_examples);
w_machine->trainers[i]->data->pushSubset(selected_examples[i],n);
// keep in mind examples not used by trainers[i]
int k=0;
for (int j=0;j<n;j++) {
if (!is_selected_examples[j])
unselected_examples[i][k++] = j;
}
n_unselected_examples[i] = k;
// train the trainer
w_machine->trainers[i]->machine->reset();
w_machine->trainers[i]->train(w_machine->trainers_measurers ? w_machine->trainers_measurers[i] : NULL);
// put back the selected_examples
w_machine->trainers[i]->data->popSubset();
w_machine->n_trainers_trained = i+1;
// if measurers is given, call the test method by fooling it
// with the number of trainers
if (measurers)
test(measurers);
}
}
Bagging::~Bagging()
{
free(is_selected_examples);
free(n_unselected_examples);
for (int i=0;i<n_trainers;i++) {
free(unselected_examples[i]);
free(selected_examples[i]);
}
free(selected_examples);
free(unselected_examples);
}
}
|
