71605439c2
git-svn-id: svn://scm.gforge.inria.fr/svnroot/paradiseo@1890 331e1502-861f-0410-8da2-ba01fb791d7f
435 lines
10 KiB
C++
435 lines
10 KiB
C++
// j'ai installé le svn :)
|
|
// re-test
|
|
|
|
/*
|
|
* <moeo2DMinHypervolumeArchive.h>
|
|
* Copyright (C) DOLPHIN Project-Team, INRIA Futurs, 2006-2007
|
|
* (C) OPAC Team, LIFL, 2002-2007
|
|
*
|
|
* Arnaud Liefooghe
|
|
* Jérémie Humeau
|
|
*
|
|
* 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 MOEO2DMINHYPERVOLUMEARCHIVE_H_
|
|
#define MOEO2DMINHYPERVOLUMEARCHIVE_H_
|
|
|
|
#include <set>
|
|
#include <climits>
|
|
|
|
|
|
template < class MOEOT >
|
|
struct comp
|
|
{
|
|
// returns a "before" b
|
|
// all objectives = min
|
|
bool operator() (const MOEOT & a, const MOEOT & b)
|
|
{
|
|
return ((a.objectiveVector()[1] < b.objectiveVector()[1]) || ((a.objectiveVector()[1] == b.objectiveVector()[1]) && (a.objectiveVector()[0] < b.objectiveVector()[0])));
|
|
}
|
|
};
|
|
|
|
|
|
/** 2D (minimization) bounded archive by hypervolume , base on a set */
|
|
template < class MOEOT >
|
|
class moeo2DMinHypervolumeArchive : public std::set<MOEOT , comp < MOEOT > >
|
|
{
|
|
public:
|
|
|
|
typedef typename MOEOT::Fitness Fitness;
|
|
typedef typename MOEOT::ObjectiveVector ObjectiveVector;
|
|
typedef typename std::set < MOEOT, comp<MOEOT> >::iterator Iterator;
|
|
|
|
using std::set < MOEOT, comp<MOEOT> > :: begin;
|
|
using std::set < MOEOT, comp<MOEOT> > :: end;
|
|
using std::set < MOEOT, comp<MOEOT> > :: insert;
|
|
using std::set < MOEOT, comp<MOEOT> > :: erase;
|
|
using std::set < MOEOT, comp<MOEOT> > :: size;
|
|
using std::set < MOEOT, comp<MOEOT> > :: upper_bound;
|
|
|
|
|
|
/**
|
|
* Ctr.
|
|
* @param _maxSize size of the archive (must be >= 2)
|
|
* @param _maxValue fitness assigned to the first and the last solution in the archive (default LONG_MAX)
|
|
*/
|
|
moeo2DMinHypervolumeArchive(unsigned int _maxSize=100, double _maxValue=LONG_MAX) : std::set < MOEOT, comp<MOEOT> > (), maxSize(_maxSize), maxValue(_maxValue)
|
|
{
|
|
maxSize = std::max((unsigned int) 2, maxSize);
|
|
}
|
|
|
|
|
|
/**
|
|
* Update the archive with a solution
|
|
* @param _moeo a solution
|
|
* @return true if _moeo has been added to the archive
|
|
*/
|
|
bool operator()(const MOEOT & _moeo)
|
|
{
|
|
//store result
|
|
bool result;
|
|
Iterator it;
|
|
|
|
//If archive is empty -> add the sol and affect its fitness value
|
|
if (size()==0)
|
|
{
|
|
result = true;
|
|
insert(_moeo);
|
|
it=begin();
|
|
fitness(it, maxValue);
|
|
}
|
|
else // test if sol can be added to the archive
|
|
{
|
|
result = insert(_moeo.objectiveVector());
|
|
if (result)
|
|
{
|
|
if(size() < maxSize){
|
|
// if yes, insert it and recompute fitness value of MOEOT and its neighbors
|
|
insert(hint,_moeo);
|
|
if(size() > 2)
|
|
{
|
|
//general case
|
|
hint--;
|
|
computeFitness(hint);
|
|
}
|
|
else
|
|
{
|
|
//archive size <= 2, fitness=maxValue for each sol
|
|
it=begin();
|
|
while(it!=end())
|
|
{
|
|
fitness(it, maxValue);
|
|
it++;
|
|
}
|
|
}
|
|
}
|
|
else{
|
|
result = filter(_moeo);
|
|
}
|
|
}
|
|
}
|
|
return result;
|
|
}
|
|
|
|
|
|
/**
|
|
* update the archive with a population
|
|
* @param _pop a pop
|
|
* @return true if at least one solution of _pop has been added to the archive
|
|
*/
|
|
bool operator()(const eoPop < MOEOT > & _pop)
|
|
{
|
|
bool result = false;
|
|
bool tmp = false;
|
|
for (unsigned int i=0; i<_pop.size(); i++)
|
|
{
|
|
std::cout << "insert " << _pop[i].objectiveVector()[0] << ", " << _pop[i].objectiveVector()[1] << std::endl;
|
|
tmp = (*this)(_pop[i]);
|
|
result = tmp || result;
|
|
}
|
|
return result;
|
|
}
|
|
|
|
|
|
/**
|
|
* Test if insertion wrt Pareto-dominance is possible, and fix 'hint' if possible
|
|
* @param _objVec the objective vector of the sol to insert
|
|
* @return true if objVec can be added to the archive wrt Pareto-dominance
|
|
*/
|
|
bool insert(const ObjectiveVector & _objVec)
|
|
{
|
|
bool result = false;
|
|
Iterator it;
|
|
double min;
|
|
// set the objVec to the empty solution
|
|
empty.objectiveVector(_objVec);
|
|
// compute the position where it would possibly be added
|
|
it = upper_bound(empty);
|
|
// compute the weigth from the previous solution
|
|
min = begin()->objectiveVector()[0];
|
|
if (it != begin())
|
|
{
|
|
it--;
|
|
min = (*it).objectiveVector()[0];
|
|
it++;
|
|
}
|
|
// if it has a better weitgh, or if it's an extreme sol, let's add it
|
|
if (it == begin() || _objVec[0]<min)
|
|
{
|
|
// remove dominated solutions
|
|
remove(it,_objVec);
|
|
// set hint to the current iterator (probably modified by "remove")
|
|
hint=it;
|
|
// set result to true
|
|
result = true;
|
|
}
|
|
return result;
|
|
}
|
|
|
|
|
|
/**
|
|
* print objective vector and fitness value of the archive
|
|
*/
|
|
void print(){
|
|
Iterator it = begin();
|
|
while(it!=end())
|
|
{
|
|
std::cout << (*it).objectiveVector()[0] << " " << (*it).objectiveVector()[1] << ", fit: " << (*it).fitness() << std::endl;
|
|
it++;
|
|
}
|
|
}
|
|
|
|
|
|
protected:
|
|
|
|
/** Size max of the archive*/
|
|
unsigned int maxSize;
|
|
/** fitness assigned to the first and the last solution in the archive */
|
|
double maxValue;
|
|
/** hint for the insertion */
|
|
Iterator hint;
|
|
|
|
/** an empty MOEOT used for checking insertion */
|
|
MOEOT empty;
|
|
|
|
|
|
/**
|
|
* set fitness
|
|
*/
|
|
void fitness(Iterator & _it, double _fitnessValue)
|
|
{
|
|
MOEOT* tmp;
|
|
tmp = (MOEOT*)&(*_it);
|
|
tmp->fitness(_fitnessValue);
|
|
}
|
|
|
|
|
|
/**
|
|
* remove solutions from the archive that are dominated by _objVec
|
|
* @param _it an iterator beginning on the first potentialy sol to remove
|
|
* @param _objVec the objective vector of the new solution
|
|
*/
|
|
void remove(Iterator & _it, const ObjectiveVector & _objVec)
|
|
{
|
|
Iterator itd;
|
|
while ((_it!=end()) && ((*_it).objectiveVector()[0] >= _objVec[0]))
|
|
{
|
|
itd = _it;
|
|
_it++;
|
|
erase(itd);
|
|
}
|
|
}
|
|
|
|
|
|
/**
|
|
* compute fitness value of a solution and its two neighbors
|
|
* @param _it refer to the solution
|
|
*/
|
|
void computeFitness(Iterator & _it)
|
|
{
|
|
Iterator tmp;
|
|
if(_it!=begin())
|
|
{
|
|
tmp=_it;
|
|
tmp--;
|
|
compute(tmp);
|
|
}
|
|
_it++;
|
|
if(_it!=end())
|
|
{
|
|
_it--;
|
|
tmp=_it;
|
|
tmp++;
|
|
compute(tmp);
|
|
}
|
|
else
|
|
{
|
|
_it--;
|
|
}
|
|
compute(_it);
|
|
}
|
|
|
|
|
|
/**
|
|
* compute fitness value of a solution
|
|
* @param _it refer to the solution
|
|
*/
|
|
void compute(Iterator & _it)
|
|
{
|
|
double x0, x1, y0, y1, fit;
|
|
if (_it==begin())
|
|
{
|
|
fitness(_it, maxValue);
|
|
}
|
|
else if ((++_it)==end())
|
|
{
|
|
_it--;
|
|
fitness(_it, maxValue);
|
|
}
|
|
else
|
|
{
|
|
_it--;
|
|
x0 = (*_it).objectiveVector()[0];
|
|
y0 = (*_it).objectiveVector()[1];
|
|
_it--;
|
|
x1 = (*_it).objectiveVector()[0];
|
|
_it++;
|
|
_it++;
|
|
y1 = (*_it).objectiveVector()[1];
|
|
_it--;
|
|
fit = (x1 - x0) * (y1 - y0);
|
|
fitness(_it, fit);
|
|
//tmp = (MOEOT*)&(*_it);
|
|
//tmp->fitness(fit);
|
|
}
|
|
}
|
|
|
|
double computeTmp(const ObjectiveVector & _objVec, int _where){
|
|
double res, tmp;
|
|
if(hint==begin() || hint==end())
|
|
res=maxValue;
|
|
else{
|
|
if(_where==0){
|
|
//on calcule la fit de celui à potentiellement inserer
|
|
res= (*hint).objectiveVector()[1] - _objVec[1];
|
|
hint--;
|
|
res*= ((*hint).objectiveVector()[0] - _objVec[0]);
|
|
hint++;
|
|
}
|
|
else if(_where <0){
|
|
// on calcule la fit de son predecesseur
|
|
res= _objVec[1] - (*hint).objectiveVector()[1];
|
|
tmp=(*hint).objectiveVector()[0];
|
|
hint--;
|
|
res*= ((*hint).objectiveVector()[0] - tmp);
|
|
hint++;
|
|
}
|
|
else{
|
|
// on calcule la fit de son successeur
|
|
res= _objVec[0] - (*hint).objectiveVector()[0];
|
|
tmp=(*hint).objectiveVector()[1];
|
|
hint++;
|
|
res*= ((*hint).objectiveVector()[1] - tmp);
|
|
hint--;
|
|
}
|
|
}
|
|
return res;
|
|
}
|
|
|
|
|
|
void filterbis(){
|
|
Iterator it, itd;
|
|
//used to find sol with minimum fitness value
|
|
double minFit = maxValue;
|
|
|
|
// remove MOEOT with the lowest fitness value while archive size > maxSize
|
|
while (size() > maxSize)
|
|
{
|
|
//find sol with minimum fitness
|
|
for(it=begin(); it!=end(); it++)
|
|
{
|
|
if(it->fitness() < minFit)
|
|
{
|
|
minFit = it->fitness();
|
|
itd = it;
|
|
}
|
|
}
|
|
//remove it and recompute fitness of its neighbors
|
|
it = itd;
|
|
it--;
|
|
erase(itd);
|
|
compute(it);
|
|
it++;
|
|
compute(it);
|
|
}
|
|
}
|
|
|
|
/**
|
|
* iteratively removes the less-contributing solution from the acrhive
|
|
*/
|
|
bool filter(const MOEOT & _moeo)
|
|
{
|
|
bool res;
|
|
double x, y, pred, succ, tmp=0;
|
|
|
|
if(hint==begin() || hint==end()){
|
|
insert(hint, _moeo);
|
|
hint--;
|
|
computeFitness(hint);
|
|
filterbis();
|
|
res=true;
|
|
}
|
|
else{
|
|
//compute fitness tmp
|
|
tmp=computeTmp(_moeo.objectiveVector(), 0);
|
|
hint--;
|
|
pred=computeTmp(_moeo.objectiveVector(), -1);
|
|
hint++;
|
|
succ=computeTmp(_moeo.objectiveVector(), 1);
|
|
if(tmp > succ || tmp>pred){
|
|
insert(hint, _moeo);
|
|
hint--;
|
|
//ici faudrait utiliser les valeurs qu'on vient de calculer pour les affecter direct (faire attention à ou on se trouve)
|
|
computeFitness(hint);
|
|
filterbis();
|
|
res=true;
|
|
}
|
|
else{
|
|
Iterator it;
|
|
double minFit = maxValue;
|
|
for(it=begin(); it!=end(); it++)
|
|
{
|
|
if(it->fitness() < minFit)
|
|
{
|
|
minFit = it->fitness();
|
|
}
|
|
}
|
|
if(tmp<=minFit){
|
|
res=false;
|
|
}
|
|
else{
|
|
//REDONDANT arranger le code
|
|
insert(hint, _moeo);
|
|
hint--;
|
|
//ici faudrait utiliser les valeurs qu'on vient de calculer pour les affecter direct (faire attention à ou on se trouve)
|
|
computeFitness(hint);
|
|
filterbis();
|
|
res=true;
|
|
}
|
|
}
|
|
}
|
|
|
|
return res;
|
|
}
|
|
|
|
};
|
|
|
|
#endif /*MOEO2DMINHYPERVOLUMEARCHIVE_H_ */
|