statscalcfitnesskozaop.cpp

非常好的进化算法EC 实现平台 可以实现多种算法 GA GP

/*
 *  Open BEAGLE
 *  Copyright (C) 2001-2005 by Christian Gagne and Marc Parizeau
 *
 *  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., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 *
 *  Contact:
 *  Laboratoire de Vision et Systemes Numeriques
 *  Departement de genie electrique et de genie informatique
 *  Universite Laval, Quebec, Canada, G1K 7P4
 *  http://vision.gel.ulaval.ca
 *
 */

/*!
 *  \file   beagle/GP/src/StatsCalcFitnessKozaOp.cpp
 *  \brief  Source code of class StatsCalcFitnessKozaOp.
 *  \author Christian Gagne
 *  \author Marc Parizeau
 *  $Revision: 1.10 $
 *  $Date: 2005/10/04 16:25:10 $
 */

#include "beagle/GP.hpp"

#include <cmath>
#include <sstream>

using namespace Beagle;


/*!
 *  \brief Construct a calculate stats operator.
 *  \param inName Name of the operator.
 */
GP::StatsCalcFitnessKozaOp::StatsCalcFitnessKozaOp(Beagle::string inName) :
  StatsCalculateOp(inName)
{ }


/*!
 *  \brief Calculate statistics of a given deme.
 *  \param outStats Evaluated statistics.
 *  \param ioDeme Deme to evalute the statistics.
 *  \param ioContext Context of the evolution.
 */
void GP::StatsCalcFitnessKozaOp::calculateStatsDeme(Beagle::Stats& outStats,
                                                    Beagle::Deme& ioDeme,
                                                    Beagle::Context& ioContext) const
{
  Beagle_StackTraceBeginM();
  outStats.clear();
  outStats.clearItems();
  outStats.addItem("processed", ioContext.getProcessedDeme());
  outStats.addItem("total-processed", ioContext.getTotalProcessedDeme());

  if(ioDeme.size() == 0) {
    outStats.setGenerationValues(string("deme")+uint2str(ioContext.getDemeIndex()+1),
                                 ioContext.getGeneration(), 0, true);

    outStats.resize(7);
    outStats[0].mId = "normalized";
    outStats[0].mAvg = 0.0;
    outStats[0].mStd = 0.0;
    outStats[0].mMax = 0.0;
    outStats[0].mMin = 0.0;

    outStats[1].mId = "adjusted";
    outStats[1].mAvg = 0.0;
    outStats[1].mStd = 0.0;
    outStats[1].mMax = 0.0;
    outStats[1].mMin = 0.0;

    outStats[2].mId = "standardized";
    outStats[2].mAvg = 0.0;
    outStats[2].mStd = 0.0;
    outStats[2].mMax = 0.0;
    outStats[2].mMin = 0.0;

    outStats[3].mId = "raw";
    outStats[3].mAvg = 0.0;
    outStats[3].mStd = 0.0;
    outStats[3].mMax = 0.0;
    outStats[3].mMin = 0.0;

    outStats[4].mId = "hits";
    outStats[4].mAvg = 0.0;
    outStats[4].mStd = 0.0;
    outStats[4].mMax = 0.0;
    outStats[4].mMin = 0.0;

    outStats[5].mId = "treedepth";
    outStats[5].mAvg = 0.0;
    outStats[5].mStd = 0.0;
    outStats[5].mMax = 0.0;
    outStats[5].mMin = 0.0;

    outStats[6].mId = "treesize";
    outStats[6].mAvg = 0.0;
    outStats[6].mStd = 0.0;
    outStats[6].mMax = 0.0;
    outStats[6].mMin = 0.0;
    return;
  }

  const GP::Deme& lGPDeme = castObjectT<const GP::Deme&>(ioDeme);
  const GP::FitnessKoza::Handle lFirstIndivFitness =
    castHandleT<GP::FitnessKoza>(lGPDeme[0]->getFitness());

  if(ioDeme.size() == 1) {
    outStats.setGenerationValues(string("deme")+uint2str(ioContext.getDemeIndex()+1),
                                 ioContext.getGeneration(), 1, true);

    outStats.resize(7);
    outStats[0].mId = "normalized";
    outStats[0].mAvg = lFirstIndivFitness->getNormalizedFitness();
    outStats[0].mStd = 0.0;
    outStats[0].mMax = lFirstIndivFitness->getNormalizedFitness();
    outStats[0].mMin = lFirstIndivFitness->getNormalizedFitness();

    outStats[1].mId = "adjusted";
    outStats[1].mAvg = lFirstIndivFitness->getAdjustedFitness();
    outStats[1].mStd = 0.0;
    outStats[1].mMax = lFirstIndivFitness->getAdjustedFitness();
    outStats[1].mMin = lFirstIndivFitness->getAdjustedFitness();

    outStats[2].mId = "standardized";
    outStats[2].mAvg = lFirstIndivFitness->getStandardizedFitness();
    outStats[2].mStd = 0.0;
    outStats[2].mMax = lFirstIndivFitness->getStandardizedFitness();
    outStats[2].mMin = lFirstIndivFitness->getStandardizedFitness();

    outStats[3].mId = "raw";
    outStats[3].mAvg = lFirstIndivFitness->getRawFitness();
    outStats[3].mStd = 0.0;
    outStats[3].mMax = lFirstIndivFitness->getRawFitness();
    outStats[3].mMin = lFirstIndivFitness->getRawFitness();

    outStats[4].mId = "hits";
    outStats[4].mAvg = lFirstIndivFitness->getHits();
    outStats[4].mStd = 0.0;
    outStats[4].mMax = lFirstIndivFitness->getHits();
    outStats[4].mMin = lFirstIndivFitness->getHits();

    outStats[5].mId = "treedepth";
    outStats[5].mAvg = lGPDeme[0]->getMaxTreeDepth();
    outStats[5].mStd = 0.0;
    outStats[5].mMax = outStats[5].mAvg;
    outStats[5].mMin = outStats[5].mAvg;

    outStats[6].mId = "treesize";
    outStats[6].mAvg = lGPDeme[0]->getTotalNodes();
    outStats[6].mStd = 0.0;
    outStats[6].mMax = outStats[6].mAvg;
    outStats[6].mMin = outStats[6].mAvg;
    return;
  }

  double lSumNrm         = (double)lFirstIndivFitness->getNormalizedFitness();
  double lPow2SumNrm     = pow2Of<double>(lSumNrm);
  float  lMaxNrm         = lFirstIndivFitness->getNormalizedFitness();
  float  lMinNrm         = lFirstIndivFitness->getNormalizedFitness();
  double lSumAdj         = (double)lFirstIndivFitness->getAdjustedFitness();
  double lPow2SumAdj     = pow2Of<double>(lSumAdj);
  float  lMaxAdj         = lFirstIndivFitness->getAdjustedFitness();
  float  lMinAdj         = lFirstIndivFitness->getAdjustedFitness();
  double lSumStd         = (double)lFirstIndivFitness->getStandardizedFitness();
  double lPow2SumStd     = pow2Of<double>(lSumStd);
  float  lMaxStd         = (double)lFirstIndivFitness->getStandardizedFitness();
  float  lMinStd         = lFirstIndivFitness->getStandardizedFitness();
  double lSumRaw         = lFirstIndivFitness->getRawFitness();
  double lPow2SumRaw     = pow2Of<double>(lSumRaw);
  float  lMaxRaw         = (double)lFirstIndivFitness->getRawFitness();
  float  lMinRaw         = lFirstIndivFitness->getRawFitness();
  double lSumHit         = lFirstIndivFitness->getHits();
  double lPow2SumHit     = pow2Of<double>(lSumHit);
  unsigned int lMaxHit   = lFirstIndivFitness->getHits();
  unsigned int lMinHit   = lFirstIndivFitness->getHits();

  unsigned int lMaxDepth = lGPDeme[0]->getMaxTreeDepth();
  unsigned int lMinDepth = lMaxDepth;
  double lSumDepth       = (double)lMaxDepth;
  double lPow2SumDepth   = pow2Of<double>(lSumDepth);
  unsigned int lMaxSize  = lGPDeme[0]->getTotalNodes();
  unsigned int lMinSize  = lMaxSize;
  double lSumSize        = (double)lMaxSize;
  double lPow2SumSize    = pow2Of<double>(lSumSize);
  
  for(unsigned int i=1; i<lGPDeme.size(); i++) {
    const GP::FitnessKoza::Handle lIndivFitness =
      castHandleT<GP::FitnessKoza>(lGPDeme[i]->getFitness());
    lSumNrm     += (double)lIndivFitness->getNormalizedFitness();
    lPow2SumNrm += pow2Of<double>((double)lIndivFitness->getNormalizedFitness());
    lSumAdj     += (double)lIndivFitness->getAdjustedFitness();
    lPow2SumAdj += pow2Of<double>((double)lIndivFitness->getAdjustedFitness());
    lSumStd     += (double)lIndivFitness->getStandardizedFitness();
    lPow2SumStd += pow2Of<double>(lIndivFitness->getStandardizedFitness());
    lSumRaw     += (double)lIndivFitness->getRawFitness();
    lPow2SumRaw += pow2Of<double>((double)lIndivFitness->getRawFitness());
    lSumHit     += (double)lIndivFitness->getHits();
    lPow2SumHit += pow2Of<double>((double)lIndivFitness->getHits());
    if(lIndivFitness->getNormalizedFitness() > lMaxNrm) {
      lMaxNrm = lIndivFitness->getNormalizedFitness();
      lMaxAdj = lIndivFitness->getAdjustedFitness();
      lMaxStd = lIndivFitness->getStandardizedFitness();
      lMaxRaw = lIndivFitness->getRawFitness();
      lMaxHit = lIndivFitness->getHits();
    }
    if(lIndivFitness->getNormalizedFitness() < lMinNrm) {
      lMinNrm = lIndivFitness->getNormalizedFitness();
      lMinAdj = lIndivFitness->getAdjustedFitness();
      lMinStd = lIndivFitness->getStandardizedFitness();
      lMinRaw = lIndivFitness->getRawFitness();
      lMinHit = lIndivFitness->getHits();
    }

    unsigned int lTmpDepth = lGPDeme[i]->getMaxTreeDepth();
    lSumDepth     += (double)lTmpDepth;
    lPow2SumDepth += pow2Of<double>((double)lTmpDepth);
    lMaxDepth     =  maxOf(lMaxDepth, lTmpDepth);
    lMinDepth     =  minOf(lMinDepth, lTmpDepth);

    unsigned int lTmpSize = lGPDeme[i]->getTotalNodes();
    lSumSize     += (double)lTmpSize;
    lPow2SumSize += pow2Of<double>((double)lTmpSize);
    lMaxSize     =  maxOf(lMaxSize, lTmpSize);
    lMinSize     =  minOf(lMinSize, lTmpSize);
  }

  float lNrmAverage  = (float)(lSumNrm / lGPDeme.size());
  float lNrmStdError =
    (float)(lPow2SumNrm - (pow2Of<double>(lSumNrm) / lGPDeme.size())) / (lGPDeme.size() - 1);
  lNrmStdError       = sqrt(lNrmStdError);

  float lAdjAverage  = (float)(lSumAdj / lGPDeme.size());
  float lAdjStdError =
    (float)(lPow2SumAdj - (pow2Of<double>(lSumAdj) / lGPDeme.size())) / (lGPDeme.size() - 1);
  lAdjStdError       = sqrt(lAdjStdError);

  float lStdAverage  = (float)(lSumStd / lGPDeme.size());
  float lStdStdError =
    (float)(lPow2SumStd - (pow2Of<double>(lSumStd) / lGPDeme.size())) / (lGPDeme.size() - 1);
  lStdStdError       = sqrt(lStdStdError);

  float lRawAverage  = (float)(lSumRaw / lGPDeme.size());
  float lRawStdError =
    (float)(lPow2SumRaw - (pow2Of<double>(lSumRaw) / lGPDeme.size())) / (lGPDeme.size() - 1);
  lRawStdError       = sqrt(lRawStdError);

  float lHitAverage  = (float)(lSumHit / lGPDeme.size());
  float lHitStdError =
    (float)(lPow2SumHit - (pow2Of<double>(lSumHit) / lGPDeme.size())) / (lGPDeme.size() - 1);
  lHitStdError       = sqrt(lHitStdError);

  float lDepthAverage  = (float)(lSumDepth / lGPDeme.size());
  float lDepthStdError =
    (float)(lPow2SumDepth - (pow2Of<double>(lSumDepth) / lGPDeme.size())) / (lGPDeme.size() - 1);
  lDepthStdError       = sqrt(lDepthStdError);

  float lSizeAverage  = (float)(lSumSize / lGPDeme.size());
  float lSizeStdError =
    (float)(lPow2SumSize - (pow2Of<double>(lSumSize) / lGPDeme.size())) / (lGPDeme.size() - 1);
  lSizeStdError       = sqrt(lSizeStdError);

  outStats.setGenerationValues(string("deme")+uint2str(ioContext.getDemeIndex()+1),
                               ioContext.getGeneration(), ioDeme.size(), true);

  outStats.resize(7);
  outStats[0].mId = "normalized";
  outStats[0].mAvg = lNrmAverage;
  outStats[0].mStd = lNrmStdError;
  outStats[0].mMax = lMaxNrm;
  outStats[0].mMin = lMinNrm;

  outStats[1].mId = "adjusted";
  outStats[1].mAvg = lAdjAverage;
  outStats[1].mStd = lAdjStdError;
  outStats[1].mMax = lMaxAdj;
  outStats[1].mMin = lMinAdj;

  outStats[2].mId = "standardized";
  outStats[2].mAvg = lStdAverage;
  outStats[2].mStd = lStdStdError;
  outStats[2].mMax = lMaxStd;
  outStats[2].mMin = lMinStd;

  outStats[3].mId = "raw";
  outStats[3].mAvg = lRawAverage;
  outStats[3].mStd = lRawStdError;
  outStats[3].mMax = lMaxRaw;
  outStats[3].mMin = lMinRaw;

  outStats[4].mId = "hits";
  outStats[4].mAvg = lHitAverage;
  outStats[4].mStd = lHitStdError;
  outStats[4].mMax = (float)lMaxHit;
  outStats[4].mMin = (float)lMinHit;

  outStats[5].mId = "treedepth";
  outStats[5].mAvg = lDepthAverage;
  outStats[5].mStd = lDepthStdError;
  outStats[5].mMax = (float)lMaxDepth;
  outStats[5].mMin = (float)lMinDepth;

  outStats[6].mId = "treesize";
  outStats[6].mAvg = lSizeAverage;
  outStats[6].mStd = lSizeStdError;
  outStats[6].mMax = (float)lMaxSize;
  outStats[6].mMin = (float)lMinSize;
  Beagle_StackTraceEndM("void GP::StatsCalcFitnessKozaOp::calculateStatsDeme(Beagle::Stats& outStats, Beagle::Deme& ioDeme, Beagle::Context& ioContext) const");
}