/usr/include/mia-2.4/mia/core/kmeans.hh is in libmia-2.4-dev 2.4.3-5.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
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 | /* -*- mia-c++ -*-
*
* This file is part of MIA - a toolbox for medical image analysis
* Copyright (c) Leipzig, Madrid 1999-2016 Gert Wollny
*
* MIA 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 3 of the License, or
* (at your option) any later version.
*
* This program 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 MIA; if not, see <http://www.gnu.org/licenses/>.
*
*/
#ifndef __mia_core_kmeans_hh
#define __mia_core_kmeans_hh
#include <vector>
#include <numeric>
#include <cmath>
#include <stdexcept>
#include <iomanip>
#include <limits>
#include <mia/core/defines.hh>
#include <mia/core/errormacro.hh>
#include <mia/core/msgstream.hh>
#include <boost/concept/requires.hpp>
#include <boost/concept_check.hpp>
NS_MIA_BEGIN
int EXPORT_CORE kmeans_get_closest_clustercenter(const std::vector<double>& classes, size_t l, double value);
template <typename InputIterator, typename OutputIterator>
bool kmeans_step(InputIterator ibegin, InputIterator iend, OutputIterator obegin,
std::vector<double>& classes, size_t l, int& biggest_class )
{
cvdebug()<< "kmeans enter: ";
for (size_t i = 0; i <= l; ++i )
cverb << std::setw(8) << classes[i]<< " ";
cverb << "\n";
biggest_class = -1;
const double convLimit = 0.005; // currently fixed
std::vector<double> sums(classes.size());
std::vector<size_t> count(classes.size());
bool conv = false;
int iter = 50;
while( iter-- && !conv) {
sort(classes.begin(), classes.end());
// assign closest cluster center
OutputIterator ob = obegin;
for (InputIterator b = ibegin; b != iend; ++b, ++ob) {
*ob = kmeans_get_closest_clustercenter(classes,l, *b);
++count[*ob];
sums[*ob] += *b;
};
// recompute cluster centers
conv = true;
size_t max_count = 0;
for (size_t i = 0; i <= l; i++) {
if (count[i]) {
double a = sums[i] / count[i];
if (a && fabs ((a - classes[i]) / a) > convLimit)
conv = false;
classes[i] = a;
if (max_count < count[i]) {
max_count = count[i];
biggest_class = i;
}
} else { // if a class is empty move it closer to neighbour
if (i == 0)
classes[i] = (classes[i] + classes[i + 1]) / 2.0;
else
classes[i] = (classes[i] + classes[i - 1]) / 2.0;
conv = false;
}
sums[i] = 0;
count[i] = 0;
};
};
cvinfo()<< "kmeans: " << l + 1 << " classes, " << 50 - iter << " iterations";
for (size_t i = 0; i <= l; ++i )
cverb << std::setw(8) << classes[i]<< " ";
cverb << "\n";
return conv;
}
template <typename InputIterator, typename OutputIterator>
bool kmeans_step_with_fixed_centers(InputIterator ibegin, InputIterator iend, OutputIterator obegin,
std::vector<double>& classes, const std::vector<bool>& fixed_center,
size_t l, int& biggest_class )
{
cvdebug()<< "kmeans enter: ";
for (size_t i = 0; i <= l; ++i )
cverb << std::setw(8) << classes[i]<< " ";
cverb << "\n";
biggest_class = -1;
const double convLimit = 0.005; // currently fixed
std::vector<double> sums(classes.size());
std::vector<size_t> count(classes.size());
bool conv = false;
int iter = 50;
while( iter-- && !conv) {
sort(classes.begin(), classes.end());
// assign closest cluster center
OutputIterator ob = obegin;
for (InputIterator b = ibegin; b != iend; ++b, ++ob) {
*ob = kmeans_get_closest_clustercenter(classes,l, *b);
++count[*ob];
sums[*ob] += *b;
};
// recompute cluster centers
conv = true;
size_t max_count = 0;
for (size_t i = 0; i <= l; i++) {
if (fixed_center[i])
continue;
if (count[i]) {
double a = sums[i] / count[i];
if (a && fabs ((a - classes[i]) / a) > convLimit)
conv = false;
classes[i] = a;
if (max_count < count[i]) {
max_count = count[i];
biggest_class = i;
}
} else { // if a class is empty move it closer to neighbour
if (i == 0)
classes[i] = (classes[i] + classes[i + 1]) / 2.0;
else
classes[i] = (classes[i] + classes[i - 1]) / 2.0;
conv = false;
}
sums[i] = 0;
count[i] = 0;
};
};
cvinfo()<< "kmeans: " << l + 1 << " classes, " << 50 - iter << " iterations";
for (size_t i = 0; i <= l; ++i )
cverb << std::setw(8) << classes[i]<< " ";
cverb << "\n";
return conv;
}
/**
\ingroup misc
Run a kmeans clustering on some input data and store the class centers and the
class membership.
\tparam InputIterator readable forward iterator,
\tparam OutputIterator writable forward iterator,
\param ibegin iterator indicating the start of the input data
\param iend iterator indicating the end of the input data, expect an STL-like handling,
i.e. iend points behind the last element to be accessed
\param obegin begin of the output range where the class membership will be stored
it is up to the caller to ensure that this range is at least as large as the input range
\param[in,out] classes at input the size of the vector indicates the number of clusters to be used
at output the vector elements contain the class centers in increasing order.
*/
template <typename InputIterator, typename OutputIterator>
BOOST_CONCEPT_REQUIRES( ((::boost::ForwardIterator<InputIterator>))
((::boost::Mutable_ForwardIterator<OutputIterator>)),
(void)
)
kmeans(InputIterator ibegin, InputIterator iend, OutputIterator obegin,
std::vector<double>& classes)
{
if (classes.size() < 2)
throw create_exception<std::invalid_argument>("kmeans: requested ", classes.size(),
"class(es), required are at least two");
const size_t nclusters = classes.size();
const double size = std::distance(ibegin, iend);
if ( size < nclusters )
throw create_exception<std::invalid_argument>("kmeans: insufficient input: want ", nclusters ,
" classes, but git only ", size, " input elements");
double sum = std::accumulate(ibegin, iend, 0.0);
// simple initialization splitting at the mean
classes[0] = sum / (size - 1);
classes[1] = sum / (size + 1);
// first run calles directly
int biggest_class = 0;
// coverity is completely off here, the 1UL is actually a class index
// and has nothing to do with the size of the type pointed to by ibegin
//
// coverity[sizeof_mismatch]
kmeans_step(ibegin, iend, obegin, classes, 1, biggest_class);
// further clustering always splits biggest class
for (size_t l = 2; l < nclusters; l++) {
const size_t pos = biggest_class > 0 ? biggest_class - 1 : biggest_class + 1;
classes[l] = 0.5 * (classes[biggest_class] + classes[pos]);
kmeans_step(ibegin, iend, obegin, classes, l, biggest_class);
};
// some post iteration until centers no longer change
for (size_t l = 1; l < 3; l++) {
if (kmeans_step(ibegin, iend, obegin, classes, nclusters - 1, biggest_class))
break;
}
}
NS_MIA_END
#endif
|