paradiseo/moeo/src/metric/moeoHyperVolumeDifferenceMetric.h

286 lines
11 KiB
C++

/*
* <moeoHyperVolumeDifferenceMetric.h>
* Copyright (C) DOLPHIN Project-Team, INRIA Futurs, 2006-2007
* (C) OPAC Team, LIFL, 2002-2007
*
* Jeremie Humeau
* Arnaud Liefooghe
*
* This software is governed by the CeCILL license under French law and
* abiding by the rules of distribution of free software. You can use,
* modify and/ or redistribute the software under the terms of the CeCILL
* license as circulated by CEA, CNRS and INRIA at the following URL
* "http://www.cecill.info".
*
* As a counterpart to the access to the source code and rights to copy,
* modify and redistribute granted by the license, users are provided only
* with a limited warranty and the software's author, the holder of the
* economic rights, and the successive licensors have only limited liability.
*
* In this respect, the user's attention is drawn to the risks associated
* with loading, using, modifying and/or developing or reproducing the
* software by the user in light of its specific status of free software,
* that may mean that it is complicated to manipulate, and that also
* therefore means that it is reserved for developers and experienced
* professionals having in-depth computer knowledge. Users are therefore
* encouraged to load and test the software's suitability as regards their
* requirements in conditions enabling the security of their systems and/or
* data to be ensured and, more generally, to use and operate it in the
* same conditions as regards security.
* The fact that you are presently reading this means that you have had
* knowledge of the CeCILL license and that you accept its terms.
*
* ParadisEO WebSite : http://paradiseo.gforge.inria.fr
* Contact: paradiseo-help@lists.gforge.inria.fr
*
*/
//-----------------------------------------------------------------------------
#ifndef MOEOHYPERVOLUMEDIFFERENCEMETRIC_H_
#define MOEOHYPERVOLUMEDIFFERENCEMETRIC_H_
#include <metric/moeoMetric.h>
#include <metric/moeoHyperVolumeMetric.h>
/**
* The contribution metric evaluates the proportion of non-dominated solutions given by a Pareto set relatively to another Pareto set
* (Meunier, Talbi, Reininger: 'A multiobjective genetic algorithm for radio network optimization', in Proc. of the 2000 Congress on Evolutionary Computation, IEEE Press, pp. 317-324)
*/
template < class ObjectiveVector >
class moeoHyperVolumeDifferenceMetric : public moeoVectorVsVectorBinaryMetric < ObjectiveVector, double >
{
public:
/**
* Constructor with a coefficient (rho)
* @param _normalize allow to normalize data (default true)
* @param _rho coefficient to determine the reference point.
*/
moeoHyperVolumeDifferenceMetric(bool _normalize=true, double _rho=1.1): normalize(_normalize), rho(_rho), ref_point(/*NULL*/){
bounds.resize(ObjectiveVector::Traits::nObjectives());
// initialize bounds in case someone does not want to use them
for (unsigned int i=0; i<ObjectiveVector::Traits::nObjectives(); i++)
{
bounds[i] = eoRealInterval(0,1);
}
}
/**
* Constructor with a reference point
* @param _normalize allow to normalize data (default true)
* @param _ref_point the reference point
*/
moeoHyperVolumeDifferenceMetric(bool _normalize/*=true*/, ObjectiveVector& _ref_point/*=NULL*/): normalize(_normalize), rho(0.0), ref_point(_ref_point){
bounds.resize(ObjectiveVector::Traits::nObjectives());
// initialize bounds in case someone does not want to use them
for (unsigned int i=0; i<ObjectiveVector::Traits::nObjectives(); i++)
{
bounds[i] = eoRealInterval(0,1);
}
}
/**
* calculates and returns the HyperVolume value of a pareto front
* @param _set1 the vector contains all objective Vector of the first pareto front
* @param _set2 the vector contains all objective Vector of the second pareto front
*/
virtual double operator()(const std::vector < ObjectiveVector > & _set1, const std::vector < ObjectiveVector > & _set2)
{
double hypervolume_set1;
double hypervolume_set2;
if(rho >= 1.0){
//determine bounds
setup(_set1, _set2);
//determine reference point
for (unsigned int i=0; i<ObjectiveVector::Traits::nObjectives(); i++){
if(normalize){
if (ObjectiveVector::Traits::minimizing(i))
ref_point[i]= rho;
else
ref_point[i]= 1-rho;
}
else{
if (ObjectiveVector::Traits::minimizing(i))
ref_point[i]= bounds[i].maximum() * rho;
else
ref_point[i]= bounds[i].maximum() * (1-rho);
}
}
//if no normalization, reinit bounds to O..1 for
if(!normalize)
for (unsigned int i=0; i<ObjectiveVector::Traits::nObjectives(); i++)
bounds[i] = eoRealInterval(0,1);
}
else if(normalize)
setup(_set1, _set2);
moeoHyperVolumeMetric <ObjectiveVector> unaryMetric(ref_point, bounds);
hypervolume_set1 = unaryMetric(_set1);
hypervolume_set2 = unaryMetric(_set2);
return hypervolume_set1 - hypervolume_set2;
}
/**
* getter on bounds
* @return bounds
*/
std::vector < eoRealInterval > getBounds(){
return bounds;
}
/**
* method calculate bounds for the normalization
* @param _set1 the vector contains all objective Vector of the first pareto front
* @param _set2 the vector contains all objective Vector of the second pareto front
*/
void setup(const std::vector < ObjectiveVector > & _set1, const std::vector < ObjectiveVector > & _set2)
{
if(_set1.size() < 1 || _set2.size() < 1) {
throw("Error in moeoHyperVolumeUnaryMetric::setup -> argument1: vector<ObjectiveVector> size must be greater than 0");
} else {
#ifndef NDEBUG
if( _set1.size() == 1 || _set2.size() == 1 ) {
eo::log << eo::warnings << "Warning in moeoHyperVolumeUnaryMetric::setup one of the pareto set contains only one point (set1.size="
<< _set1.size() << ", set2.size=" << _set2.size() << ")"
<< std::endl;
}
#endif
typename ObjectiveVector::Type min, max;
unsigned int nbObj=ObjectiveVector::Traits::nObjectives();
bounds.resize(nbObj);
for (unsigned int i=0; i<nbObj; i++){
min = _set1[0][i];
max = _set1[0][i];
for (unsigned int j=1; j<_set1.size(); j++){
min = std::min(min, _set1[j][i]);
max = std::max(max, _set1[j][i]);
}
for (unsigned int j=0; j<_set2.size(); j++){
min = std::min(min, _set2[j][i]);
max = std::max(max, _set2[j][i]);
}
if( min == max ) {
bounds[i] = eoRealInterval(min-tiny(), max+tiny());
} else {
bounds[i] = eoRealInterval(min, max);
}
}
}
}
protected:
/**
* Returns a very small value that can be used to avoid extreme cases (where the min bound == the max bound)
*/
static double tiny()
{
return 1e-6;
}
protected:
/*boolean indicates if data must be normalized or not*/
bool normalize;
double rho;
/*vectors contains bounds for normalization*/
std::vector < eoRealInterval > bounds;
ObjectiveVector ref_point;
};
template<class ObjectiveVector>
class moeoDualHyperVolumeDifferenceMetric : public moeoHyperVolumeDifferenceMetric<ObjectiveVector>
{
protected:
using moeoHyperVolumeDifferenceMetric<ObjectiveVector>::rho;
using moeoHyperVolumeDifferenceMetric<ObjectiveVector>::normalize;
using moeoHyperVolumeDifferenceMetric<ObjectiveVector>::ref_point;
using moeoHyperVolumeDifferenceMetric<ObjectiveVector>::bounds;
public:
typedef typename ObjectiveVector::Type Type;
moeoDualHyperVolumeDifferenceMetric( bool _normalize=true, double _rho=1.1)
: moeoHyperVolumeDifferenceMetric<ObjectiveVector>(_normalize, _rho)
{
}
moeoDualHyperVolumeDifferenceMetric( bool _normalize/*=true*/, ObjectiveVector& _ref_point/*=NULL*/ )
: moeoHyperVolumeDifferenceMetric<ObjectiveVector>( _normalize, _ref_point )
{
}
/**
* calculates and returns the HyperVolume value of a pareto front
* @param _set1 the vector contains all objective Vector of the first pareto front
* @param _set2 the vector contains all objective Vector of the second pareto front
*/
virtual double operator()(const std::vector < ObjectiveVector > & _set1, const std::vector < ObjectiveVector > & _set2)
{
#ifndef NDEBUG
// the two sets must be homogeneous in feasibility
assert( _set1.size() > 0 );
for( unsigned int i=1; i<_set1.size(); ++i ) {
assert( _set1[i].is_feasible() == _set1[0].is_feasible() );
}
assert( _set2.size() > 0 );
for( unsigned int i=1; i<_set2.size(); ++i ) {
assert( _set2[i].is_feasible() == _set2[0].is_feasible() );
}
// and they must have the same feasibility
assert( _set1[0].is_feasible() == _set2[0].is_feasible() );
#endif
bool feasible = _set1[0].is_feasible();
double hypervolume_set1;
double hypervolume_set2;
if(rho >= 1.0){
//determine bounds
setup(_set1, _set2);
//determine reference point
for (unsigned int i=0; i<ObjectiveVector::Traits::nObjectives(); i++){
if(normalize){
if (ObjectiveVector::Traits::minimizing(i))
ref_point[i]= Type(rho, feasible);
else
ref_point[i]= Type(1-rho, feasible);
}
else{
if (ObjectiveVector::Traits::minimizing(i))
ref_point[i]= Type(bounds[i].maximum() * rho, feasible);
else
ref_point[i]= Type(bounds[i].maximum() * (1-rho), feasible);
}
}
//if no normalization, reinit bounds to O..1 for
if(!normalize)
for (unsigned int i=0; i<ObjectiveVector::Traits::nObjectives(); i++)
bounds[i] = eoRealInterval(0,1);
}
else if(normalize)
setup(_set1, _set2);
moeoHyperVolumeMetric <ObjectiveVector> unaryMetric(ref_point, bounds);
hypervolume_set1 = unaryMetric(_set1);
hypervolume_set2 = unaryMetric(_set2);
return hypervolume_set1 - hypervolume_set2;
}
};
#endif /*MOEOHYPERVOLUMEMETRIC_H_*/