source: branches/peters_vector/lib/classifier/ConsensusInputRanker.cc @ 469

// $Id: ConsensusInputRanker.cc 469 2005-12-19 14:58:29Z peter $


#include <c++_tools/classifier/ConsensusInputRanker.h>

#include <c++_tools/classifier/CrossSplitting.h>
#include <c++_tools/classifier/DataView.h>
#include <c++_tools/classifier/InputRanker.h>
#include <c++_tools/classifier/Target.h>
#include <c++_tools/statistics/utility.h>
#include <c++_tools/utility/stl_utility.h>
#include <c++_tools/gslapi/matrix.h>

#include <utility>
#include <vector>

namespace theplu {
namespace classifier {  

  ConsensusInputRanker::ConsensusInputRanker
  (const DataView& data, const Target& target,
   statistics::Score& score_object, CrossSplitting& sampler,
   const size_t n)
    : nof_rankers_(n)

  {
    for (size_t i=0; i<n; i++){
      std::vector<size_t> index=sampler.next();
      input_rankers_.push_back(InputRanker(DataView(data,index), 
                                           Target(target,index), 
                                           score_object)  );
    }
    // Sorting with respect to median rank
    std::vector<std::pair<size_t,double> > medians(data.rows());
    for (size_t i=0; i<data.rows(); i++){ 
      std::vector<size_t> ranks(nof_rankers_);
      for (size_t j=0; j<nof_rankers_; j++) 
        ranks[j]=input_rankers_[j].rank(i);
      
      medians[i].first = i;
      medians[i].second = statistics::median(ranks);
    }
    
    //sort medians and assign id_ and rank_
    sort(medians.begin(), medians.end(), 
         utility::pair_value_compare<size_t, double>());
    id_.resize(data.rows());
    rank_.resize(data.rows());
    for (size_t i=0; i<data.rows(); i++){
      id_[i]=medians[i].first;
      rank_[id_[i]]=i;            
    }

  }

  ConsensusInputRanker::ConsensusInputRanker
  (const DataView& data, const Target& target, const DataView& weight,
   statistics::Score& score_object, CrossSplitting& sampler,const size_t n)
    : nof_rankers_(n)
  {

    for (size_t i=0; i<n; i++){
      std::vector<size_t> index=sampler.next();
      input_rankers_.push_back(InputRanker(DataView(data,index),
                                           Target(target,index),
                                           score_object,
                                           DataView(weight,index)));
    }
    
    // Sorting with respect to median rank
    std::vector<std::pair<size_t,double> > median(data.rows());
    for (size_t i=0; i<data.rows(); i++){ 
      std::vector<size_t> ranks(nof_rankers_);
      for (size_t j=0; j<nof_rankers_; j++) 
        ranks[j]=input_rankers_[j].rank(i);
      median[i].first = i;
      median[i].second = statistics::median(ranks);
    }
    
    //sort medians and assign id_ and rank_
    sort(median.begin(), median.end(), 
         utility::pair_value_compare<size_t, double>());
    id_.resize(data.rows());
    rank_.resize(data.rows());
    for (size_t i=0; i<data.rows(); i++){
      id_[i]=median[i].first;
      rank_[id_[i]]=i;            
    }

  }
    

}} // of namespace classifier and namespace theplu