/usr/include/vigra/random_forest/rf_preprocessing.hxx is in libvigraimpex-dev 1.10.0+git20160211.167be93+dfsg-2+b5.
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 | /************************************************************************/
/* */
/* Copyright 2008-2009 by Ullrich Koethe and Rahul Nair */
/* */
/* This file is part of the VIGRA computer vision library. */
/* The VIGRA Website is */
/* http://hci.iwr.uni-heidelberg.de/vigra/ */
/* Please direct questions, bug reports, and contributions to */
/* ullrich.koethe@iwr.uni-heidelberg.de or */
/* vigra@informatik.uni-hamburg.de */
/* */
/* Permission is hereby granted, free of charge, to any person */
/* obtaining a copy of this software and associated documentation */
/* files (the "Software"), to deal in the Software without */
/* restriction, including without limitation the rights to use, */
/* copy, modify, merge, publish, distribute, sublicense, and/or */
/* sell copies of the Software, and to permit persons to whom the */
/* Software is furnished to do so, subject to the following */
/* conditions: */
/* */
/* The above copyright notice and this permission notice shall be */
/* included in all copies or substantial portions of the */
/* Software. */
/* */
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND */
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES */
/* OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND */
/* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT */
/* HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, */
/* WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING */
/* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR */
/* OTHER DEALINGS IN THE SOFTWARE. */
/* */
/************************************************************************/
#ifndef VIGRA_RF_PREPROCESSING_HXX
#define VIGRA_RF_PREPROCESSING_HXX
#include <limits>
#include <vigra/mathutil.hxx>
#include "rf_common.hxx"
namespace vigra
{
/** Class used while preprocessing (currently used only during learn)
*
* This class is internally used by the Random Forest learn function.
* Different split functors may need to process the data in different manners
* (i.e., regression labels that should not be touched and classification
* labels that must be converted into a integral format)
*
* This Class only exists in specialized versions, where the Tag class is
* fixed.
*
* The Tag class is determined by Splitfunctor::Preprocessor_t . Currently
* it can either be ClassificationTag or RegressionTag. look At the
* RegressionTag specialisation for the basic interface if you ever happen
* to care.... - or need some sort of vague new preprocessor.
* new preprocessor ( Soft labels or whatever)
*/
template<class Tag, class LabelType, class T1, class C1, class T2, class C2>
class Processor;
namespace detail
{
/* Common helper function used in all Processors.
* This function analyses the options struct and calculates the real
* values needed for the current problem (data)
*/
template<class T>
void fill_external_parameters(RandomForestOptions const & options,
ProblemSpec<T> & ext_param)
{
// set correct value for mtry
switch(options.mtry_switch_)
{
case RF_SQRT:
ext_param.actual_mtry_ =
int(std::floor(
std::sqrt(double(ext_param.column_count_))
+ 0.5));
break;
case RF_LOG:
// this is in Breimans original paper
ext_param.actual_mtry_ =
int(1+(std::log(double(ext_param.column_count_))
/std::log(2.0)));
break;
case RF_FUNCTION:
ext_param.actual_mtry_ =
options.mtry_func_(ext_param.column_count_);
break;
case RF_ALL:
ext_param.actual_mtry_ = ext_param.column_count_;
break;
default:
ext_param.actual_mtry_ =
options.mtry_;
}
// set correct value for msample
switch(options.training_set_calc_switch_)
{
case RF_CONST:
ext_param.actual_msample_ =
options.training_set_size_;
break;
case RF_PROPORTIONAL:
ext_param.actual_msample_ =
static_cast<int>(std::ceil(options.training_set_proportion_ *
ext_param.row_count_));
break;
case RF_FUNCTION:
ext_param.actual_msample_ =
options.training_set_func_(ext_param.row_count_);
break;
default:
vigra_precondition(1!= 1, "unexpected error");
}
}
/* Returns true if MultiArray contains NaNs
*/
template<unsigned int N, class T, class C>
bool contains_nan(MultiArrayView<N, T, C> const & in)
{
typedef typename MultiArrayView<N, T, C>::const_iterator Iter;
Iter i = in.begin(), end = in.end();
for(; i != end; ++i)
if(isnan(NumericTraits<T>::toRealPromote(*i)))
return true;
return false;
}
/* Returns true if MultiArray contains Infs
*/
template<unsigned int N, class T, class C>
bool contains_inf(MultiArrayView<N, T, C> const & in)
{
if(!std::numeric_limits<T>::has_infinity)
return false;
typedef typename MultiArrayView<N, T, C>::const_iterator Iter;
Iter i = in.begin(), end = in.end();
for(; i != end; ++i)
if(abs(*i) == std::numeric_limits<T>::infinity())
return true;
return false;
}
} // namespace detail
/** Preprocessor used during Classification
*
* This class converts the labels int Integral labels which are used by the
* standard split functor to address memory in the node objects.
*/
template<class LabelType, class T1, class C1, class T2, class C2>
class Processor<ClassificationTag, LabelType, T1, C1, T2, C2>
{
public:
typedef Int32 LabelInt;
typedef MultiArrayView<2, T1, C1> Feature_t;
typedef MultiArray<2, T1> FeatureWithMemory_t;
typedef MultiArrayView<2,LabelInt> Label_t;
MultiArrayView<2, T1, C1>const & features_;
MultiArray<2, LabelInt> intLabels_;
MultiArrayView<2, LabelInt> strata_;
template<class T>
Processor(MultiArrayView<2, T1, C1>const & features,
MultiArrayView<2, T2, C2>const & response,
RandomForestOptions &options,
ProblemSpec<T> &ext_param)
:
features_( features) // do not touch the features.
{
vigra_precondition(!detail::contains_nan(features), "RandomForest(): Feature matrix "
"contains NaNs");
vigra_precondition(!detail::contains_nan(response), "RandomForest(): Response "
"contains NaNs");
vigra_precondition(!detail::contains_inf(features), "RandomForest(): Feature matrix "
"contains inf");
vigra_precondition(!detail::contains_inf(response), "RandomForest(): Response "
"contains inf");
// set some of the problem specific parameters
ext_param.column_count_ = features.shape(1);
ext_param.row_count_ = features.shape(0);
ext_param.problem_type_ = CLASSIFICATION;
ext_param.used_ = true;
intLabels_.reshape(response.shape());
//get the class labels
if(ext_param.class_count_ == 0)
{
// fill up a map with the current labels and then create the
// integral labels.
std::set<T2> labelToInt;
for(MultiArrayIndex k = 0; k < features.shape(0); ++k)
labelToInt.insert(response(k,0));
std::vector<T2> tmp_(labelToInt.begin(), labelToInt.end());
ext_param.classes_(tmp_.begin(), tmp_.end());
}
for(MultiArrayIndex k = 0; k < features.shape(0); ++k)
{
if(std::find(ext_param.classes.begin(), ext_param.classes.end(), response(k,0)) == ext_param.classes.end())
{
throw std::runtime_error("RandomForest(): invalid label in training data.");
}
else
intLabels_(k, 0) = std::find(ext_param.classes.begin(), ext_param.classes.end(), response(k,0))
- ext_param.classes.begin();
}
// set class weights
if(ext_param.class_weights_.size() == 0)
{
ArrayVector<T2>
tmp(static_cast<std::size_t>(ext_param.class_count_),
NumericTraits<T2>::one());
ext_param.class_weights(tmp.begin(), tmp.end());
}
// set mtry and msample
detail::fill_external_parameters(options, ext_param);
// set strata
strata_ = intLabels_;
}
/** Access the processed features
*/
MultiArrayView<2, T1, C1>const & features()
{
return features_;
}
/** Access processed labels
*/
MultiArrayView<2, LabelInt> response()
{
return MultiArrayView<2, LabelInt>(intLabels_);
}
/** Access processed strata
*/
ArrayVectorView < LabelInt> strata()
{
return ArrayVectorView<LabelInt>(intLabels_.size(), intLabels_.data());
}
/** Access strata fraction sized - not used currently
*/
ArrayVectorView< double> strata_prob()
{
return ArrayVectorView< double>();
}
};
/** Regression Preprocessor - This basically does not do anything with the
* data.
*/
template<class LabelType, class T1, class C1, class T2, class C2>
class Processor<RegressionTag,LabelType, T1, C1, T2, C2>
{
public:
// only views are created - no data copied.
MultiArrayView<2, T1, C1> features_;
MultiArrayView<2, T2, C2> response_;
RandomForestOptions const & options_;
ProblemSpec<LabelType> const &
ext_param_;
// will only be filled if needed
MultiArray<2, int> strata_;
bool strata_filled;
// copy the views.
template<class T>
Processor( MultiArrayView<2, T1, C1> features,
MultiArrayView<2, T2, C2> response,
RandomForestOptions const & options,
ProblemSpec<T>& ext_param)
:
features_(features),
response_(response),
options_(options),
ext_param_(ext_param)
{
// set some of the problem specific parameters
ext_param.column_count_ = features.shape(1);
ext_param.row_count_ = features.shape(0);
ext_param.problem_type_ = REGRESSION;
ext_param.used_ = true;
detail::fill_external_parameters(options, ext_param);
vigra_precondition(!detail::contains_nan(features), "Processor(): Feature Matrix "
"Contains NaNs");
vigra_precondition(!detail::contains_nan(response), "Processor(): Response "
"Contains NaNs");
vigra_precondition(!detail::contains_inf(features), "Processor(): Feature Matrix "
"Contains inf");
vigra_precondition(!detail::contains_inf(response), "Processor(): Response "
"Contains inf");
strata_ = MultiArray<2, int> (MultiArrayShape<2>::type(response_.shape(0), 1));
ext_param.response_size_ = response.shape(1);
ext_param.class_count_ = response_.shape(1);
std::vector<T2> tmp_(ext_param.class_count_, 0);
ext_param.classes_(tmp_.begin(), tmp_.end());
}
/** access preprocessed features
*/
MultiArrayView<2, T1, C1> & features()
{
return features_;
}
/** access preprocessed response
*/
MultiArrayView<2, T2, C2> & response()
{
return response_;
}
/** access strata - this is not used currently
*/
MultiArray<2, int> & strata()
{
return strata_;
}
};
}
#endif //VIGRA_RF_PREPROCESSING_HXX
|