outputinfo.cpp

MultiBoost 是c++实现的多类adaboost酸法。与传统的adaboost算法主要解决二类分类问题不同

/*
* This file is part of MultiBoost, a multi-class 
* AdaBoost learner/classifier
*
* Copyright (C) 2005-2006 Norman Casagrande
* For informations write to [e-mail]
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
*
*/

#include <limits>

#include "IO/OutputInfo.h"
#include "IO/ClassMappings.h"
#include "WeakLearners/BaseLearner.h"

namespace MultiBoost {

// -------------------------------------------------------------------------

OutputInfo::OutputInfo(const string& outputInfoFile)
{
   // open the stream
   _outStream.open(outputInfoFile.c_str());

   // is it really open?
   if ( !_outStream.is_open() )
   {
      cerr << "ERROR: cannot open the output steam (<" 
           << outputInfoFile << ">) for the step-by-step info!" << endl;
      exit(1);
   }
}

// -------------------------------------------------------------------------

void OutputInfo::outputIteration(int t)
{ 
   _outStream << t; // just output t
}

// -------------------------------------------------------------------------

void OutputInfo::outputError(InputData* pData, BaseLearner* pWeakHypothesis)
{
   const int numClasses = ClassMappings::getNumClasses();
   const int numExamples = pData->getNumExamples();

   if ( _gTableMap.find(pData) == _gTableMap.end() )
   {
      // if it's the first time it sees this data
      // it creates and initializes a new table

      table& g = _gTableMap[pData];
      g.resize(numExamples);

      for ( int i = 0; i < numExamples; ++i )
         g[i].resize(numClasses, 0);
   }
   
   table& g = _gTableMap[pData];
   
   int numWrongs = 0;

   // Compute the training error
   for (int i = 0; i < numExamples; ++i)
   {
      // the class with the highest vote
      int maxClassIdx = -1;

      // the vote of the winning class
      double maxClass = -numeric_limits<double>::max();

      for (int l = 0; l < numClasses; ++l)
      {     
         // building the strong learner
         g[i][l] += pWeakHypothesis->getAlpha() * // alpha
                    pWeakHypothesis->classify( pData, i, l ); // h_l(x)

         // get the winner class
         if (g[i][l] > maxClass)
         {
            maxClass = g[i][l];
            maxClassIdx = l;
         }

      }

      // if the winner class is not the actual class, then it is
      // an error
      if (maxClassIdx != pData->getClass(i))
         ++numWrongs;
   }

   // The error must be bounded between 0 and 1
   _outStream << '\t' << (double)(numWrongs)/(double)(numExamples);
}

// -------------------------------------------------------------------------

void OutputInfo::outputMargins(InputData* pData, BaseLearner* pWeakHypothesis)
{
   const int numClasses = ClassMappings::getNumClasses();
   const int numExamples = pData->getNumExamples();

   if ( _margins.find(pData) == _margins.end() )
   {
      // if it's the first time it sees this data
      // it creates and initializes a new table

      table& margins = _margins[pData];
      margins.resize(numExamples);

      for ( int i = 0; i < numExamples; ++i )
         margins[i].resize(numClasses, 0);
   }

   // Same for the sums of alpha. If it is the first time it
   // sees this data, it initialize the sums of the alpha for it
   if ( _alphaSums.find(pData) == _alphaSums.end() )
      _alphaSums[pData] = 0;

   table& margins = _margins[pData];

   double minMargin = numeric_limits<double>::max();
   double belowZeroMargin = 0;                        

   for (int i = 0; i < numExamples; ++i)
   {
      for (int l = 0; l < numClasses; ++l)
      {
         // hy = +1 if the classification it is correct, -1 otherwise
         double hy = pWeakHypothesis->classify(pData, i, l) * // h_l(x_i)
                     pData->getBinaryClass(i, l); // y_i

         // compute the margin
         margins[i][l] += pWeakHypothesis->getAlpha() * hy;

         // gets the margin below zero
         if ( margins[i][l] < 0 )
         {
            if (l == pData->getClass(i))
               belowZeroMargin += ( 1.0 / static_cast<double>(2*numExamples) );
            else
               belowZeroMargin += ( 1.0 / static_cast<double>(2*numExamples * (numClasses - 1)) );
         }        

         // get the minimum margin among classes and examples
         if (margins[i][l] < minMargin)
            minMargin = margins[i][l];
      }
   }

   // compute the sums of the alphas for normalization
   _alphaSums[pData] += pWeakHypothesis->getAlpha();

   _outStream << '\t' << minMargin/_alphaSums[pData] << "\t" // minimum margin
              << belowZeroMargin; // margins that are below zero
}

// -------------------------------------------------------------------------

void OutputInfo::outputEdge(InputData* pData, BaseLearner* pWeakHypothesis)
{
   const int numClasses = ClassMappings::getNumClasses();
   const int numExamples = pData->getNumExamples();

   double gamma = 0; // the edge

   for (int i = 0; i < numExamples; ++i)
   {
      for (int l = 0; l < numClasses; ++l)
      {      
         // hy = +1 if the classification it is correct, -1 otherwise
         double hy = pWeakHypothesis->classify(pData, i, l) * // h_l(x_i)
                     pData->getBinaryClass(i, l); // y_i

         double w = pData->getWeight(i, l);

         gamma += w * hy;
      }
   }

   _outStream << '\t' << gamma; // edge

}

// -------------------------------------------------------------------------

} // end of namespace MultiBoost