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
|
// 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 "KNN.h"
namespace Torch {
KNN::KNN(DataSet* data_,int K_)
{
data = data_;
distances = NULL;
indices = NULL;
setK(K_);
n_outputs = data->n_targets;
addToList(&outputs, n_outputs, (real *)xalloc(sizeof(real)*n_outputs));
n_real_examples = 0;
real_examples = NULL;
}
void KNN::reset()
{
data->setExample(0);
n_real_examples = data->n_examples;
real_examples = (int*)xrealloc(real_examples,n_real_examples*sizeof(int));
for (int i=0;i<data->n_examples;i++) {
real_examples[i] = data->selected_examples[i];
}
}
void KNN::setK(int K_)
{
K = K_;
distances = (real *)xrealloc(distances,K*sizeof(real));
indices = (int *)xrealloc(indices,K*sizeof(int));
}
void KNN::forward(List* inputs)
{
// keep current example in order to restore at the end
int current_example = data->current_example;
real* inputs_ptr = (real*)inputs->ptr;
// verify that n_examples > K
int old_K = K;
if (n_real_examples < K)
K = n_real_examples;
// initialization of distances to big values;
for (int i=0;i<K;i++) {
distances[i] = 1e35;
indices[i] = -1;
}
// compute the K nearest neighbords
// for each vector in data
int* i_ptr = real_examples;
for (int i=0;i<n_real_examples;i++) {
data->setRealExample(*i_ptr++);
// calculate euclidean distance between example and current vector
real dist = 0;
real *x = inputs_ptr;
real *datax = (real *)data->inputs->ptr;
for(int j = 0; j < data->n_inputs; j++) {
real diff = *x++ - *datax++;
dist += diff*diff;
}
// eventually add current vector to K nearest neighbors
if (dist < distances[K-1]) {
// find insertion point
real* bptr = distances;
real* eptr = distances + K - 1;
real* mptr = bptr + (eptr - bptr) / 2;
do {
if (dist < *mptr)
eptr = mptr;
else
bptr = mptr + 1;
mptr = bptr + (eptr - bptr) / 2;
} while (mptr < eptr);
// insert the point by shifting all subsequent distances
eptr = distances + K - 1;
bptr = eptr - 1;
int* eptr_idx = indices + K - 1;
int* bptr_idx = eptr_idx - 1;
while (eptr > mptr) {
*eptr-- = *bptr--; /* distances */
*eptr_idx-- = *bptr_idx--; /* indices */
}
*mptr = dist;
indices[mptr - distances] = data->current_example;
}
}
// give an answer as the mean of the answers of the KNNs
// initialize outputs to null
real* out = (real*)outputs->ptr;
for (int j=0;j<n_outputs;j++)
*out++ = 0;
for (int i=0;i<K;i++) {
out = (real*)outputs->ptr;
data->setRealExample(indices[i]);
real *targ = (real*)data->targets;
for (int j=0;j<n_outputs;j++)
*out++ += *targ++;
}
out = (real*)outputs->ptr;
for (int j=0;j<n_outputs;j++)
*out++ /= (real)K;
// in case K was modified
K = old_K;
// restore current_example
data->setRealExample(current_example);
}
KNN::~KNN()
{
freeList(&outputs,true);
free(distances);
free(indices);
free(real_examples);
}
}
|
