Ajout du Metropolis-Hasting LS, du samplinf MH fitness cloud

git-svn-id: svn://scm.gforge.inria.fr/svnroot/paradiseo@1792 331e1502-861f-0410-8da2-ba01fb791d7f
This commit is contained in:
verel 2010-05-06 17:03:38 +00:00
commit 71fa51d3b5
6 changed files with 222 additions and 2 deletions

View file

@ -0,0 +1,110 @@
/*
<moMetropolisHasting.h>
Copyright (C) DOLPHIN Project-Team, INRIA Lille - Nord Europe, 2006-2010
Sebastien Verel, Arnaud Liefooghe, Jeremie Humeau
This software is governed by the CeCILL license under French law and
abiding by the rules of distribution of free software. You can ue,
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".
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 _moMetropolisHasting_h
#define _moMetropolisHasting_h
#include <algo/moLocalSearch.h>
#include <explorer/moMetropolisHastingExplorer.h>
#include <continuator/moTrueContinuator.h>
#include <eval/moEval.h>
#include <eoEvalFunc.h>
/********************************************************
* Metropolis-Hasting local search
* Only the symetric case is considered when Q(x,y) = Q(y,x)
* Fitness must be > 0
*
* At each iteration,
* one of the random solution in the neighborhood is selected
* if the selected neighbor have higher or equal fitness than the current solution
* then the solution is replaced by the selected neighbor
* if a random number from [0,1] is lower than fitness(neighbor) / fitness(solution)
* then the solution is replaced by the selected neighbor
* the algorithm stops when the number of iterations is too large
********************************************************/
template<class Neighbor>
class moMetropolisHasting: public moLocalSearch<Neighbor>
{
public:
typedef typename Neighbor::EOT EOT;
typedef moNeighborhood<Neighbor> Neighborhood ;
/**
* Simple constructor of the Metropolis-Hasting
* @param _neighborhood the neighborhood
* @param _fullEval the full evaluation function
* @param _eval neighbor's evaluation function
* @param _nbStep maximum step to do
*/
moMetropolisHasting(Neighborhood& _neighborhood, eoEvalFunc<EOT>& _fullEval, moEval<Neighbor>& _eval, unsigned int _nbStep):
moLocalSearch<Neighbor>(explorer, trueCont, _fullEval),
explorer(_neighborhood, _eval, defaultNeighborComp, defaultSolNeighborComp, _nbStep)
{}
/**
* Simple constructor of the Metropolis-Hasting
* @param _neighborhood the neighborhood
* @param _fullEval the full evaluation function
* @param _eval neighbor's evaluation function
* @param _nbStep maximum step to do
* @param _cont an external continuator
*/
moMetropolisHasting(Neighborhood& _neighborhood, eoEvalFunc<EOT>& _fullEval, moEval<Neighbor>& _eval, unsigned int _nbStep, moContinuator<Neighbor>& _cont):
moLocalSearch<Neighbor>(explorer, _cont, _fullEval),
explorer(_neighborhood, _eval, defaultNeighborComp, defaultSolNeighborComp, _nbStep)
{}
/**
* Simple constructor of the Metropolis-Hasting
* @param _neighborhood the neighborhood
* @param _fullEval the full evaluation function
* @param _eval neighbor's evaluation function
* @param _nbStep maximum step to do
* @param _cont an external continuator
* @param _compN a neighbor vs neighbor comparator
* @param _compSN a solution vs neighbor comparator
*/
moMetropolisHasting(Neighborhood& _neighborhood, eoEvalFunc<EOT>& _fullEval, moEval<Neighbor>& _eval, unsigned int _nbStep, moContinuator<Neighbor>& _cont, moNeighborComparator<Neighbor>& _compN, moSolNeighborComparator<Neighbor>& _compSN):
moLocalSearch<Neighbor>(explorer, _cont, _fullEval),
explorer(_neighborhood, _eval, _compN, _compSN, _nbStep)
{}
private:
// always true continuator
moTrueContinuator<Neighbor> trueCont;
// compare the fitness values of neighbors: true is strictly greater
moNeighborComparator<Neighbor> defaultNeighborComp;
// compare the fitness values of the solution and the neighbor: true if strictly greater
moSolNeighborComparator<Neighbor> defaultSolNeighborComp;
// the explorer of the HC with neutral move (equals fitness move)
moMetropolisHastingExplorer<Neighbor> explorer;
};
#endif

View file

@ -57,6 +57,7 @@ public:
* @param _neighborhood the neighborhood
* @param _fullEval the full evaluation function
* @param _eval neighbor's evaluation function
* @param _nbStep maximum step to do
*/
moNeutralHC(Neighborhood& _neighborhood, eoEvalFunc<EOT>& _fullEval, moEval<Neighbor>& _eval, unsigned int _nbStep):
moLocalSearch<Neighbor>(explorer, trueCont, _fullEval),
@ -68,6 +69,7 @@ public:
* @param _neighborhood the neighborhood
* @param _fullEval the full evaluation function
* @param _eval neighbor's evaluation function
* @param _nbStep maximum step to do
* @param _cont an external continuator
*/
moNeutralHC(Neighborhood& _neighborhood, eoEvalFunc<EOT>& _fullEval, moEval<Neighbor>& _eval, unsigned int _nbStep, moContinuator<Neighbor>& _cont):
@ -80,6 +82,7 @@ public:
* @param _neighborhood the neighborhood
* @param _fullEval the full evaluation function
* @param _eval neighbor's evaluation function
* @param _nbStep maximum step to do
* @param _cont an external continuator
* @param _compN a neighbor vs neighbor comparator
* @param _compSN a solution vs neighbor comparator

View file

@ -37,6 +37,7 @@
#include <algo/moLocalSearch.h>
#include <algo/moRandomSearch.h>
#include <algo/moMetropolisHasting.h>
#include <algo/moSA.h>
#include <algo/moSimpleHC.h>
#include <algo/moFirstImprHC.h>
@ -143,7 +144,8 @@
#include <sampling/moHillClimberSampling.h>
#include <sampling/moFDCsampling.h>
#include <sampling/moNeutralDegreeSampling.h>
#include <sampling/moFitnessCouldSampling.h>
#include <sampling/moFitnessCloudSampling.h>
#include <sampling/moMHFitnessCloudSampling.h>
#include <problems/bitString/moBitNeighbor.h>
#include <problems/eval/moOneMaxIncrEval.h>

View file

@ -0,0 +1,100 @@
/*
<moMHFitnessCloudSampling.h>
Copyright (C) DOLPHIN Project-Team, INRIA Lille - Nord Europe, 2006-2010
Sebastien Verel, Arnaud Liefooghe, Jeremie 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 moMHFitnessCloudSampling_h
#define moMHFitnessCloudSampling_h
#include <eoInit.h>
#include <neighborhood/moNeighborhood.h>
#include <eval/moEval.h>
#include <eoEvalFunc.h>
#include <algo/moMetropolisHasting.h>
#include <continuator/moFitnessStat.h>
#include <continuator/moNeighborFitnessStat.h>
#include <sampling/moSampling.h>
/**
* To compute an estimation of the fitness cloud,
* i.e. the scatter plot of solution fitness versus neighbor fitness:
* Here solution are sampled with Metropolis-Hasting method
*
* Sample the fitness of solutions from Metropolis-Hasting sampling
* and the fitness of one random neighbor
*
* The values are collected during the random search
*
*/
template <class Neighbor>
class moMHFitnessCloudSampling : public moSampling<Neighbor>
{
public:
typedef typename Neighbor::EOT EOT ;
using moSampling<Neighbor>::localSearch;
/**
* Default Constructor
* @param _init initialisation method of the solution
* @param _neighborhood neighborhood to get one random neighbor (supposed to be random neighborhood)
* @param _fullEval Fitness function, full evaluation function
* @param _eval neighbor evaluation, incremental evaluation function
* @param _nbStep Number of step of the MH sampling
*/
moMHFitnessCloudSampling(eoInit<EOT> & _init,
moNeighborhood<Neighbor> & _neighborhood,
eoEvalFunc<EOT>& _fullEval,
moEval<Neighbor>& _eval,
unsigned int _nbStep) :
moSampling<Neighbor>(_init, * new moMetropolisHasting<Neighbor>(_neighborhood, _fullEval, _eval, _nbStep), fitnessStat),
neighborFitnessStat(_neighborhood, _eval)
{
add(neighborFitnessStat);
}
/**
* default destructor
*/
~moMHFitnessCloudSampling() {
// delete the pointer on the local search which has been constructed in the constructor
delete &localSearch;
}
protected:
moFitnessStat<EOT> fitnessStat;
moNeighborFitnessStat< Neighbor > neighborFitnessStat;
};
#endif

View file

@ -62,6 +62,7 @@ public:
* @param _init initialisation method of the solution
* @param _localSearch local search to sample the search space
* @param _stat statistic to compute during the search
* @param _monitoring the statistic is saved into the monitor if true
*/
template <class ValueType>
moSampling(eoInit<EOT> & _init, moLocalSearch<Neighbor> & _localSearch, moStat<EOT,ValueType> & _stat, bool _monitoring = true) : init(_init), localSearch(_localSearch), continuator(_localSearch.getContinuator())
@ -85,6 +86,7 @@ public:
/**
* Add a statistic
* @param _stat another statistic to compute during the search
* @param _monitoring the statistic is saved into the monitor if true
*/
template< class ValueType >
void add(moStat<EOT, ValueType> & _stat, bool _monitoring = true) {

View file

@ -38,6 +38,7 @@ using namespace std;
//-----------------------------------------------------------------------------
// the sampling class
#include <sampling/moFitnessCloudSampling.h>
#include <sampling/moMHFitnessCloudSampling.h>
// Declaration of types
//-----------------------------------------------------------------------------
@ -163,7 +164,9 @@ void main_function(int argc, char **argv)
// - fitness function
// - neighbor evaluation
// - number of solutions to sample
moFitnessCloudSampling<Neighbor> sampling(random, neighborhood, fullEval, neighborEval, nbSol);
//moFitnessCloudSampling<Neighbor> sampling(random, neighborhood, fullEval, neighborEval, nbSol);
moMHFitnessCloudSampling<Neighbor> sampling(random, neighborhood, fullEval, neighborEval, nbSol);
/* =========================================================
*