35#ifndef VIGRA_RF_ALGORITHM_HXX
36#define VIGRA_RF_ALGORITHM_HXX
67 int columnCount = std::distance(b, e);
68 int rowCount =
in.shape(0);
71 for(Iter iter = b; iter != e; ++iter, ++
ii)
73 columnVector(
out,
ii) = columnVector(
in, *iter);
101 template<
class Feature_t,
class Response_t>
110 return oob.oob_breiman;
126 typedef std::vector<int> FeatureList_t;
127 typedef std::vector<double> ErrorList_t;
128 typedef FeatureList_t::iterator Pivot_t;
167 vigra_precondition(std::distance(b, e) ==
static_cast<std::ptrdiff_t
>(
selected.size()),
168 "Number of features in ranking != number of features matrix");
224 std::map<typename ResponseT::value_type, int>
res_map;
225 std::vector<int>
cts;
227 for(
int ii = 0;
ii < response.shape(0); ++
ii)
239 /
double(response.shape(0));
294template<
class FeatureT,
class ResponseT,
class ErrorRateCallBack>
300 VariableSelectionResult::FeatureList_t & selected = result.
selected;
301 VariableSelectionResult::ErrorList_t & errors = result.
errors;
302 VariableSelectionResult::Pivot_t & pivot = result.pivot;
303 int featureCount = features.shape(1);
309 vigra_precondition(
static_cast<int>(selected.size()) == featureCount,
310 "forward_selection(): Number of features in Feature "
311 "matrix and number of features in previously used "
312 "result struct mismatch!");
320 VariableSelectionResult::Pivot_t next = pivot;
323 std::swap(*pivot, *next);
325 detail::choose( features,
331 std::swap(*pivot, *next);
337 std::advance(next, pos);
338 std::swap(*pivot, *next);
339 errors[std::distance(selected.begin(), pivot)] =
current_errors[pos];
342 std::cerr <<
"Choosing " << *pivot <<
" at error of " <<
current_errors[pos] << std::endl;
348template<
class FeatureT,
class ResponseT>
351 VariableSelectionResult & result)
396template<
class FeatureT,
class ResponseT,
class ErrorRateCallBack>
402 int featureCount = features.shape(1);
403 VariableSelectionResult::FeatureList_t & selected = result.
selected;
404 VariableSelectionResult::ErrorList_t & errors = result.
errors;
405 VariableSelectionResult::Pivot_t & pivot = result.pivot;
412 vigra_precondition(
static_cast<int>(selected.size()) == featureCount,
413 "backward_elimination(): Number of features in Feature "
414 "matrix and number of features in previously used "
415 "result struct mismatch!");
417 pivot = selected.end() - 1;
422 VariableSelectionResult::Pivot_t next = selected.begin();
426 std::swap(*pivot, *next);
428 detail::choose( features,
434 std::swap(*pivot, *next);
439 next = selected.begin();
440 std::advance(next, pos);
441 std::swap(*pivot, *next);
443 errors[std::distance(selected.begin(), pivot)-1] =
current_errors[pos];
447 std::cerr <<
"Eliminating " << *pivot <<
" at error of " <<
current_errors[pos] << std::endl;
453template<
class FeatureT,
class ResponseT>
456 VariableSelectionResult & result)
493template<
class FeatureT,
class ResponseT,
class ErrorRateCallBack>
499 VariableSelectionResult::FeatureList_t & selected = result.
selected;
500 VariableSelectionResult::ErrorList_t & errors = result.
errors;
501 VariableSelectionResult::Pivot_t & iter = result.pivot;
502 int featureCount = features.shape(1);
508 vigra_precondition(
static_cast<int>(selected.size()) == featureCount,
509 "forward_selection(): Number of features in Feature "
510 "matrix and number of features in previously used "
511 "result struct mismatch!");
515 for(; iter != selected.end(); ++iter)
519 detail::choose( features,
524 errors[std::distance(selected.begin(), iter)] =
error;
526 std::copy(selected.begin(), iter+1, std::ostream_iterator<int>(std::cerr,
", "));
527 std::cerr <<
"Choosing " << *(iter+1) <<
" at error of " <<
error << std::endl;
533template<
class FeatureT,
class ResponseT>
536 VariableSelectionResult & result)
543enum ClusterLeafTypes{c_Leaf = 95, c_Node = 99};
559 ClusterNode(
int nCol,
560 BT::T_Container_type & topology,
561 BT::P_Container_type & split_param)
562 : BT(nCol + 5, 5,topology, split_param)
572 ClusterNode( BT::T_Container_type
const & topology,
573 BT::P_Container_type
const & split_param,
575 :
NodeBase(5 , 5,topology, split_param, n)
581 ClusterNode( BT & node_)
586 BT::parameter_size_ += 0;
592 void set_index(
int in)
619 : parent(p), level(
l), addr(a), infm(
in)
648 double dist_func(
double a,
double b)
650 return std::min(a, b);
656 template<
class Functor>
660 std::vector<int>
stack;
661 stack.push_back(begin_addr);
662 while(!
stack.empty())
664 ClusterNode node(topology_, parameters_,
stack.back());
668 if(node.columns_size() != 1)
670 stack.push_back(node.child(0));
671 stack.push_back(node.child(1));
679 template<
class Functor>
683 std::queue<HC_Entry>
queue;
689 while(!
queue.empty())
691 level =
queue.front().level;
692 parent =
queue.front().parent;
693 addr =
queue.front().addr;
694 infm =
queue.front().infm;
695 ClusterNode node(topology_, parameters_,
queue.front().addr);
699 parnt = ClusterNode(topology_, parameters_, parent);
703 if(node.columns_size() != 1)
713 void save(std::string
file, std::string
prefix)
718 Shp(topology_.
size(),1),
722 Shp(parameters_.
size(), 1),
723 parameters_.data()));
733 template<
class T,
class C>
737 std::vector<std::pair<int, int> > addr;
739 for(
int ii = 0;
ii < distance.shape(0); ++
ii)
741 addr.push_back(std::make_pair(topology_.
size(),
ii));
742 ClusterNode
leaf(1, topology_, parameters_);
743 leaf.set_index(index);
745 leaf.columns_begin()[0] =
ii;
748 while(addr.size() != 1)
754 (addr.begin()+
jj_min)->second);
755 for(
unsigned int ii = 0;
ii < addr.
size(); ++
ii)
760 (addr.begin()+
jj_min)->second)
761 >
dist((addr.begin()+
ii)->second,
762 (addr.begin()+
jj)->second))
765 (addr.begin()+
jj)->second);
779 (addr.begin() +
ii_min)->first);
782 (addr.begin() +
jj_min)->first);
793 (addr.begin() +
ii_min)->first);
796 (addr.begin() +
jj_min)->first);
797 parent.parameters_begin()[0] =
min_dist;
798 parent.set_index(index);
802 parent.columns_begin());
806 if(*parent.columns_begin() == *
firstChild.columns_begin())
808 parent.child(0) = (addr.begin()+
ii_min)->first;
809 parent.child(1) = (addr.begin()+
jj_min)->first;
813 addr.erase(addr.begin()+
jj_min);
817 parent.child(1) = (addr.begin()+
ii_min)->first;
818 parent.child(0) = (addr.begin()+
jj_min)->first;
822 addr.erase(addr.begin()+
ii_min);
830 double bla = dist_func(
833 (addr.begin()+
jj)->second));
836 (addr.begin()+
jj)->second) =
bla;
837 dist((addr.begin()+
jj)->second,
859 bool operator()(Node& node)
872template<
class Iter,
class DT>
887 template<
class Feat_T,
class Label_T>
896 :tmp_mem_(_spl(a, b).size(),
feats.shape(1)),
899 feats_(_spl(a,b).size(),
feats.shape(1)),
900 labels_(_spl(a,b).size(),1),
906 copy_splice(_spl(a,b),
907 _spl(
feats.shape(1)),
910 copy_splice(_spl(a,b),
911 _spl(
labls.shape(1)),
917 bool operator()(Node& node)
921 int class_count = perm_imp.shape(1) - 1;
923 for(
int kk = 0;
kk < nPerm; ++
kk)
926 for(
int ii = 0;
ii < rowCount(feats_); ++
ii)
928 int index = random.uniformInt(rowCount(feats_) -
ii) +
ii;
929 for(
int jj = 0;
jj < node.columns_size(); ++
jj)
931 if(node.columns_begin()[
jj] != feats_.shape(1))
932 tmp_mem_(
ii, node.columns_begin()[
jj])
933 = tmp_mem_(index, node.columns_begin()[
jj]);
937 for(
int ii = 0;
ii < rowCount(tmp_mem_); ++
ii)
940 .predictLabel(rowVector(tmp_mem_,
ii))
944 ++perm_imp(index,labels_(
ii, 0));
946 ++perm_imp(index, class_count);
976 void save(std::string
file, std::string
prefix)
984 bool operator()(Node& node)
986 for(
int ii = 0;
ii < node.columns_size(); ++
ii)
1001 bool operator()(
Nde &
cur,
int ,
Nde parent,
bool )
1004 cur.status() = std::min(parent.status(),
cur.status());
1031 std::ofstream graphviz;
1036 std::string
const gz)
1037 :features_(features), labels_(labels),
1038 graphviz(
gz.c_str(), std::ios::out)
1040 graphviz <<
"digraph G\n{\n node [shape=\"record\"]";
1044 graphviz <<
"\n}\n";
1049 bool operator()(
Nde &
cur,
int ,
Nde parent,
bool )
1051 graphviz <<
"node" <<
cur.index() <<
" [style=\"filled\"][label = \" #Feats: "<<
cur.columns_size() <<
"\\n";
1052 graphviz <<
" status: " <<
cur.status() <<
"\\n";
1053 for(
int kk = 0;
kk <
cur.columns_size(); ++
kk)
1055 graphviz <<
cur.columns_begin()[
kk] <<
" ";
1059 graphviz <<
"\"] [color = \"" <<
cur.status() <<
" 1.000 1.000\"];\n";
1061 graphviz <<
"\"node" << parent.index() <<
"\" -> \"node" <<
cur.index() <<
"\";\n";
1081 int repetition_count_;
1087 void save(std::string filename, std::string
prefix)
1107 template<
class RF,
class PR>
1110 Int32 const class_count = rf.ext_param_.class_count_;
1111 Int32 const column_count = rf.ext_param_.column_count_+1;
1132 template<
class RF,
class PR,
class SM,
class ST>
1136 Int32 column_count = rf.ext_param_.column_count_ +1;
1137 Int32 class_count = rf.ext_param_.class_count_;
1141 typename PR::Feature_t & features
1142 =
const_cast<typename PR::Feature_t &
>(
pr.features());
1146 ArrayVector<Int32>::iterator
1149 if(rf.ext_param_.actual_msample_ <
pr.features().shape(0)- 10000)
1153 for(
int ii = 0;
ii <
pr.features().shape(0); ++
ii)
1154 indices.push_back(
ii); ;
1155 std::random_device rd;
1156 std::mt19937 g(rd());
1157 std::shuffle(indices.
begin(), indices.
end(), g);
1158 for(
int ii = 0;
ii < rf.ext_param_.row_count_; ++
ii)
1160 if(!
sm.is_used()[indices[
ii]] &&
cts[
pr.response()(indices[
ii], 0)] < 3000)
1163 ++
cts[
pr.response()(indices[
ii], 0)];
1169 for(
int ii = 0;
ii < rf.ext_param_.row_count_; ++
ii)
1170 if(!
sm.is_used()[
ii])
1189 .predictLabel(rowVector(features, *iter))
1190 ==
pr.response()(*iter, 0))
1226 template<
class RF,
class PR,
class SM,
class ST>
1234 template<
class RF,
class PR>
1274template<
class FeatureT,
class ResponseT>
1282 opt.tree_count(100);
1283 if(features.shape(0) > 40000)
1284 opt.samples_per_tree(20000).use_stratification(RF_EQUAL);
1290 RF.learn(features, response,
1292 distance =
missc.distance;
1319template<
class FeatureT,
class ResponseT>
1329template<
class Array1,
class Vector1>
1332 std::map<double, int>
mymap;
1335 for(std::map<double, int>::reverse_iterator iter =
mymap.rbegin(); iter!=
mymap.rend(); ++iter)
1337 out.push_back(iter->second);
void reshape(const difference_type &shape)
Definition multi_array.hxx:2863
Topology_type column_data() const
Definition rf_nodeproxy.hxx:159
INT & typeID()
Definition rf_nodeproxy.hxx:136
NodeBase()
Definition rf_nodeproxy.hxx:237
Parameter_type parameters_begin() const
Definition rf_nodeproxy.hxx:207
Class for a single RGB value.
Definition rgbvalue.hxx:128
Options object for the random forest.
Definition rf_common.hxx:171
size_type size() const
Definition tinyvector.hxx:913
iterator end()
Definition tinyvector.hxx:864
iterator begin()
Definition tinyvector.hxx:861
Class for fixed size vectors.
Definition tinyvector.hxx:1008
Definition rf_algorithm.hxx:1069
MultiArray< 2, double > cluster_importance_
Definition rf_algorithm.hxx:1077
MultiArray< 2, int > variables
Definition rf_algorithm.hxx:1074
void visit_at_end(RF &rf, PR &)
Definition rf_algorithm.hxx:1235
void visit_after_tree(RF &rf, PR &pr, SM &sm, ST &st, int index)
Definition rf_algorithm.hxx:1227
MultiArray< 2, double > cluster_stdev_
Definition rf_algorithm.hxx:1080
void after_tree_ip_impl(RF &rf, PR &pr, SM &sm, ST &, int index)
Definition rf_algorithm.hxx:1133
void visit_at_beginning(RF const &rf, PR const &)
Definition rf_algorithm.hxx:1108
Definition rf_algorithm.hxx:998
Definition rf_algorithm.hxx:1026
Definition rf_algorithm.hxx:965
MultiArrayView< 2, int > variables
Definition rf_algorithm.hxx:970
Definition rf_algorithm.hxx:640
void iterate(Functor &tester)
Definition rf_algorithm.hxx:657
void cluster(MultiArrayView< 2, T, C > distance)
Definition rf_algorithm.hxx:734
void breadth_first_traversal(Functor &tester)
Definition rf_algorithm.hxx:680
Definition rf_algorithm.hxx:849
NormalizeStatus(double m)
Definition rf_algorithm.hxx:855
Definition rf_algorithm.hxx:874
Definition rf_algorithm.hxx:85
double operator()(Feature_t const &features, Response_t const &response)
Definition rf_algorithm.hxx:102
RFErrorCallback(RandomForestOptions opt=RandomForestOptions())
Definition rf_algorithm.hxx:94
Definition rf_algorithm.hxx:118
double no_features
Definition rf_algorithm.hxx:152
ErrorList_t errors
Definition rf_algorithm.hxx:147
FeatureList_t selected
Definition rf_algorithm.hxx:134
bool init(FeatureT const &all_features, ResponseT const &response, ErrorRateCallBack errorcallback)
Definition rf_algorithm.hxx:206
Definition rf_visitors.hxx:1499
Definition rf_visitors.hxx:865
Definition rf_visitors.hxx:1463
Definition rf_visitors.hxx:103
void backward_elimination(FeatureT const &features, ResponseT const &response, VariableSelectionResult &result, ErrorRateCallBack errorcallback)
Definition rf_algorithm.hxx:397
void rank_selection(FeatureT const &features, ResponseT const &response, VariableSelectionResult &result, ErrorRateCallBack errorcallback)
Definition rf_algorithm.hxx:494
void forward_selection(FeatureT const &features, ResponseT const &response, VariableSelectionResult &result, ErrorRateCallBack errorcallback)
Definition rf_algorithm.hxx:295
void cluster_permutation_importance(FeatureT const &features, ResponseT const &response, HClustering &linkage, MultiArray< 2, double > &distance)
Definition rf_algorithm.hxx:1275
detail::VisitorNode< A > create_visitor(A &a)
Definition rf_visitors.hxx:345
void writeHDF5(...)
Store array data in an HDF5 file.
detail::SelectIntegerType< 32, detail::SignedIntTypes >::type Int32
32-bit signed int
Definition sized_int.hxx:175
Definition metaprogramming.hxx:123
Definition rf_algorithm.hxx:613