nojhan/paradiseo
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Merge branch 'master' into trikiSA

Browse source
This commit is contained in:
LPTK 2013-06-13 14:57:30 +02:00
commit 8c4e91584f
13 changed files with 653 additions and 281 deletions
Download patch file Download diff file Expand all files Collapse all files

111
edo/src/edoTransform.h Normal file
View file

@ -0,0 +1,111 @@
/*
The Evolving Distribution Objects framework (EDO) is a template-based,
ANSI-C++ evolutionary computation library which helps you to write your
own estimation of distribution algorithms.
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 Street, Fifth Floor, Boston, MA 02110-1301 USA
Copyright (C) 2013 Thales group
*/
/*
Authors:
Johann Dréo <johann.dreo@thalesgroup.com>
*/
#ifndef _edoTransform_h
#define _edoTransform_h
#include <eo> // eoTransform
/** @defgroup Wrappers
*
* Wrappers to interact with other parts of the framework
*/
/** Abstract base class for wrapping an estimator and a sampler as an eoTransform
*
* @ingroup Wrappers
*/
template<class D>
class edoTransform : public eoTransform< eoPop<typename D::EOType>& >
{
public:
typedef typename D::EOType EOType;
edoTransform( edoEstimator<D> & estimator, edoSampler<D> & sampler ) :
_estimator(estimator), _sampler(sampler)
{}
virtual void operator()( eoPop<EOType> & pop ) = 0;
protected:
edoEstimator<D> & _estimator;
edoSampler<D> & _sampler;
};
/** Wrapping an estimator and a sampler as an eoTransform.
*
* @ingroup Wrappers
*/
template<typename D>
class edoTransformAdaptive : public edoTransform<D>
{
public:
typedef typename D::EOType EOType;
edoTransformAdaptive( D & distrib, edoEstimator<D> & estimator, edoSampler<D> & sampler )
: _distrib(distrib), _estimator(estimator), _sampler(sampler)
{}
virtual void operator()( eoPop<EOType> & pop )
{
_distrib = _estimator( pop );
pop.clear();
for( unsigned int i = 0; i < pop.size(); ++i ) {
pop.push_back( _sampler(_distrib) );
}
}
protected:
D & _distrib;
edoEstimator<D> & _estimator;
edoSampler<D> & _sampler;
};
/** Wrapping an estimator and a sampler as an eoTransform,
* the distribution is created at instanciation and replaced at each call.
*
* @ingroup Wrappers
*/
template<typename D>
class edoTransformStateless : public edoTransformAdaptive<D>
{
public:
typedef typename D::EOType EOType;
edoTransformStateless( edoEstimator<D> & estimator, edoSampler<D> & sampler )
: edoTransformAdaptive<D>( *(new D), estimator, sampler )
{}
~edoTransformStateless()
{
// delete the temporary distrib allocated in constructor
delete &(this->_distrib);
}
};
#endif // !_edoTransform_h

167
eo/src/eoDualFitness.h
View file

@ -74,6 +74,25 @@ protected:
//! Flag that marks if the individual is feasible
bool _is_feasible;
/** Flag to prevent partial initialization
*
* The reason behind the use of this flag is a bit complicated.
* Normally, we would not want to allow initialization on a scalar.
* But in MOEO, this would necessitate to re-implement most of the
* operator computing metrics, as they expect generic scalars.
*
* As this would be too much work, we use derived metric classes and
* overload them so that they initialize dual fitnesses with the
* feasibility flag. But the compiler still must compile the base
* methods, that use the scalar interface.
*
* Thus, eoDualFitness has a scalar interface, but this flag add a
* security against partial initialization. In DEBUG mode, asserts
* will fail if the feasibility has not been explicitly initialized
* at runtime.
*/
bool _feasible_init;
public:
//! Empty initialization
@ -82,42 +101,46 @@ public:
*/
eoDualFitness() :
_value(0.0),
_is_feasible(false)
_is_feasible(false),
_feasible_init(false)
{}
//! Initialization with only the value, the fitness will be unfeasible.
/*!
* WARNING: this is what is used when you initialize a new fitness from a double.
* Unfeasible by default
* If you use this interface, you MUST set the feasibility BEFORE
* asking for it or the value. Or else, an assert will fail in debug mode.
*/
template<class T>
eoDualFitness( T value ) :
_value(value),
_is_feasible(false)
_is_feasible(false),
_feasible_init(false)
{
assert( _value == 0 );
}
//! Copy constructor
eoDualFitness(const eoDualFitness& other) :
_value(other._value),
_is_feasible(other._is_feasible)
_is_feasible(other._is_feasible),
_feasible_init(true)
{}
//! Constructor from explicit value/feasibility
eoDualFitness(const BaseType& v, const bool& is_feasible) :
_value(v),
_is_feasible(is_feasible)
_is_feasible(is_feasible),
_feasible_init(true)
{}
//! From a std::pair (first element is the value, second is the feasibility)
eoDualFitness(const std::pair<BaseType,bool>& dual) :
_value(dual.first),
_is_feasible(dual.second)
_is_feasible(dual.second),
_feasible_init(true)
{}
// FIXME is it a good idea to include implicit conversion here?
/** Conversion operator: it permits to use a fitness instance as its scalar
* type, if needed. For example, this is possible:
* eoDualFitness<double,std::less<double> > fit;
@ -129,11 +152,20 @@ public:
inline bool is_feasible() const
{
assert( _feasible_init );
return _is_feasible;
}
//! Explicitly set the feasibility. Useful if you have used previously the instantiation on a single scalar.
inline void is_feasible( bool feasible )
{
this->is_feasible( feasible );
this->_feasible_init = true;
}
inline BaseType value() const
{
assert( _feasible_init );
return _value;
}
@ -141,7 +173,7 @@ public:
eoDualFitness& operator=( const std::pair<BaseType, bool>& v )
{
this->_value = v.first;
this->_is_feasible = v.second;
this->is_feasible( v.second );
return *this;
}
@ -151,21 +183,20 @@ public:
{
if (this != &other) {
this->_value = other._value;
this->_is_feasible = other._is_feasible;
this->is_feasible( other.is_feasible() );
}
return *this;
}
/*
//! Copy operator from a scalar
template<class T>
eoDualFitness& operator=(const T v)
{
this->_value = v;
this->_is_feasible = false;
this->_feasible_init = false;
return *this;
}
*/
//! Comparison that separate feasible individuals from unfeasible ones. Feasible are always better
/*!
@ -178,11 +209,11 @@ public:
// am I better (less, by default) than the other ?
// if I'm feasible and the other is not
if( this->_is_feasible && !other._is_feasible ) {
if( this->is_feasible() && !other.is_feasible() ) {
// no, the other has a better fitness
return false;
} else if( !this->_is_feasible && other._is_feasible ) {
} else if( !this->is_feasible() && other.is_feasible() ) {
// yes, a feasible fitness is always better than an unfeasible one
return true;
@ -322,7 +353,7 @@ public:
friend
std::ostream& operator<<( std::ostream& os, const eoDualFitness<BaseType,Compare> & fitness )
{
os << fitness._value << " " << fitness._is_feasible;
os << fitness._value << " " << fitness.is_feasible();
return os;
}
@ -337,7 +368,7 @@ public:
is >> feasible;
fitness._value = value;
fitness._is_feasible = feasible;
fitness.is_feasible( feasible );
return is;
}
};
@ -355,35 +386,24 @@ template< class EOT>
bool eoIsFeasible ( const EOT & sol ) { return sol.fitness().is_feasible(); }
/** Embed two eoStat and call the first one on the feasible individuals and
* the second one on the unfeasible ones, merge the two resulting value in
* a string, separated by a given marker.
/** Separate the population into two: one with only feasible individuals, the other with unfeasible ones.
*/
//template<class EOT, class T>
template<class EOT, class EOSTAT>
class eoDualStatSwitch : public eoStat< EOT, std::string >
template<class EOT>
class eoDualPopSplit : public eoUF<const eoPop<EOT>&, void>
{
protected:
eoPop<EOT> _pop_feasible;
eoPop<EOT> _pop_unfeasible;
public:
using eoStat<EOT,std::string>::value;
// eoDualStatSwitch( eoStat<EOT,T> & stat_feasible, eoStat<EOT,T> & stat_unfeasible, std::string sep=" " ) :
eoDualStatSwitch( EOSTAT & stat_feasible, EOSTAT & stat_unfeasible, std::string sep=" " ) :
eoStat<EOT,std::string>(
"?"+sep+"?",
stat_feasible.longName()+sep+stat_unfeasible.longName()
),
_stat_feasible(stat_feasible),
_stat_unfeasible(stat_unfeasible),
_sep(sep)
{ }
virtual void operator()( const eoPop<EOT> & pop )
//! Split the pop and keep them in members
void operator()( const eoPop<EOT>& pop )
{
eoPop<EOT> pop_feasible;
pop_feasible.reserve(pop.size());
_pop_feasible.clear();
_pop_feasible.reserve(pop.size());
eoPop<EOT> pop_unfeasible;
pop_unfeasible.reserve(pop.size());
_pop_unfeasible.clear();
_pop_unfeasible.reserve(pop.size());
for( typename eoPop<EOT>::const_iterator ieot=pop.begin(), iend=pop.end(); ieot!=iend; ++ieot ) {
/*
@ -392,28 +412,71 @@ public:
}
*/
if( ieot->fitness().is_feasible() ) {
pop_feasible.push_back( *ieot );
_pop_feasible.push_back( *ieot );
} else {
pop_unfeasible.push_back( *ieot );
_pop_unfeasible.push_back( *ieot );
}
}
}
_stat_feasible( pop_feasible );
_stat_unfeasible( pop_unfeasible );
//! Merge feasible and unfeasible populations into a new one
eoPop<EOT> merge() const
{
eoPop<EOT> merged;
merged.reserve( _pop_feasible.size() + _pop_unfeasible.size() );
std::copy( _pop_feasible.begin(), _pop_feasible.end(), std::back_inserter<eoPop<EOT> >(merged) );
std::copy( _pop_unfeasible.begin(), _pop_unfeasible.end(), std::back_inserter<eoPop<EOT> >(merged) );
return merged;
}
eoPop<EOT>& feasible() { return _pop_feasible; }
eoPop<EOT>& unfeasible() { return _pop_unfeasible; }
};
/** Embed two eoStat and call the first one on the feasible individuals and
* the second one on the unfeasible ones, merge the two resulting value in
* a string, separated by a given marker.
*/
template<class EOSTAT>
class eoDualStatSwitch : public eoStat< typename EOSTAT::EOType, std::string >
{
public:
typedef typename EOSTAT::EOType EOType;
protected:
EOSTAT & _stat_feasible;
EOSTAT & _stat_unfeasible;
std::string _sep;
eoDualPopSplit<EOType> _pop_split;
public:
using eoStat<EOType,std::string>::value;
eoDualStatSwitch( EOSTAT & stat_feasible, EOSTAT & stat_unfeasible, std::string sep=" " ) :
eoStat<EOType,std::string>(
"?"+sep+"?",
stat_feasible.longName()+sep+stat_unfeasible.longName()
),
_stat_feasible(stat_feasible),
_stat_unfeasible(stat_unfeasible),
_sep(sep)
{ }
virtual void operator()( const eoPop<EOType> & pop )
{
// create two separated pop in this operator
_pop_split( pop );
_stat_feasible( _pop_split.feasible() );
_stat_unfeasible( _pop_split.unfeasible() );
std::ostringstream out;
out << _stat_feasible.value() << _sep << _stat_unfeasible.value();
value() = out.str();
}
protected:
// eoStat<EOT,T> & _stat_feasible;
// eoStat<EOT,T> & _stat_unfeasible;
EOSTAT & _stat_feasible;
EOSTAT & _stat_unfeasible;
std::string _sep;
};
/** @} */

53
eo/src/utils/eoStat.h
View file

@ -40,6 +40,7 @@ Contact: http://eodev.sourceforge.net
#include <eoPop.h>
#include <utils/eoMonitor.h>
//#include <utils/eoCheckPoint.h>
#include <utils/eoLogger.h>
/** @defgroup Stats Statistics computation
*
@ -84,6 +85,7 @@ template <class EOT, class T>
class eoStat : public eoValueParam<T>, public eoStatBase<EOT>
{
public:
typedef EOT EOType;
eoStat(T _value, std::string _description)
: eoValueParam<T>(_value, _description)
@ -120,6 +122,7 @@ template <class EOT, class ParamType>
class eoSortedStat : public eoSortedStatBase<EOT>, public eoValueParam<ParamType>
{
public :
typedef EOT EOType;
eoSortedStat(ParamType _value, std::string _desc) : eoValueParam<ParamType>(_value, _desc) {}
virtual std::string className(void) const { return "eoSortedStat"; }
@ -472,6 +475,51 @@ public :
};
*/
//! A robust measure of the mass (generally used to compute the median). Do not alter the given pop.
template<class EOT>
class eoNthElementStat : public eoStat< EOT, typename EOT::Fitness >
{
protected:
int _nth;
double _ratio;
public:
using eoStat<EOT, typename EOT::Fitness>::value;
eoNthElementStat( int nth = 0, std::string description = "NthElement")
: eoStat<EOT,typename EOT::Fitness>( 0.0, description ), _nth(nth), _ratio(-1.0)
{}
eoNthElementStat( double ratio = 0.5, std::string description = "Median" )
: eoStat<EOT,typename EOT::Fitness>( 0.0, description ), _nth(-1), _ratio(ratio)
{}
virtual void operator()( const eoPop<EOT> & _pop )
{
if( _nth == -1 ) { // asked for a ratio
_nth = static_cast<int>( std::floor(_pop.size() * _ratio) );
} else {
assert( _ratio == -1 ); // asked for a position
}
if( _pop.size() == 0 ) {
//FIXME how to implement value() = 0 ?
eo::log << eo::warnings << "Called " << className() << " on an empty pop, value unchanged" << std::endl;
} else {
eoPop<EOT> pop = _pop; // copy, thus no sorting of the original pop
std::nth_element( pop.begin(), pop.begin()+_nth, pop.end() );
value() = pop[_nth].fitness();
}
}
virtual std::string className(void) const { return "eoNthElementStat"; }
};
/** @example t-eoIQRStat.cpp
*/
//! A robust measure of dispersion (also called midspread or middle fifty) that is the difference between the third and the first quartile.
template<class EOT>
@ -480,12 +528,13 @@ class eoInterquartileRangeStat : public eoStat< EOT, typename EOT::Fitness >
public:
using eoStat<EOT, typename EOT::Fitness>::value;
eoInterquartileRangeStat( typename EOT::Fitness start, std::string description = "IQR" ) : eoStat<EOT,typename EOT::Fitness>( start, description ) {}
eoInterquartileRangeStat( std::string description = "IQR" ) : eoStat<EOT,typename EOT::Fitness>( 0.0, description ) {}
virtual void operator()( const eoPop<EOT> & _pop )
{
if( _pop.size() == 0 ) {
// how to implement value() = 0 ?
//FIXME how to implement value() = 0 ?
eo::log << eo::warnings << "Called " << className() << " on an empty pop, value unchanged" << std::endl;
} else {
eoPop<EOT> pop = _pop;

6
moeo/src/algo/moeoIBEA.h
View file

@ -160,7 +160,7 @@ public:
* @param _op variation operators
* @param _fitnessAssignment fitness assignment
*/
moeoIBEA (eoContinue < MOEOT > & _continuator, eoPopEvalFunc < MOEOT > & _popEval, eoGenOp < MOEOT > & _op, moeoExpBinaryIndicatorBasedFitnessAssignment < MOEOT >& _fitnessAssignment) :
moeoIBEA (eoContinue < MOEOT > & _continuator, eoPopEvalFunc < MOEOT > & _popEval, eoGenOp < MOEOT > & _op, moeoBinaryIndicatorBasedFitnessAssignment < MOEOT >& _fitnessAssignment) :
defaultGenContinuator(0), continuator(_continuator), eval(defaultEval), defaultPopEval(eval), popEval(_popEval), select(2),
selectMany(select,0.0), selectTransform(defaultSelect, defaultTransform), defaultSGAGenOp(defaultQuadOp, 1.0, defaultMonOp, 1.0), genBreed(select, _op), breed(genBreed), default_fitnessAssignment(NULL), fitnessAssignment(_fitnessAssignment), replace(fitnessAssignment, diversityAssignment)
{}
@ -173,7 +173,7 @@ public:
* @param _op variation operators
* @param _fitnessAssignment fitness assignment
*/
moeoIBEA (eoContinue < MOEOT > & _continuator, eoEvalFunc < MOEOT > & _eval, eoGenOp < MOEOT > & _op, moeoExpBinaryIndicatorBasedFitnessAssignment < MOEOT >& _fitnessAssignment) :
moeoIBEA (eoContinue < MOEOT > & _continuator, eoEvalFunc < MOEOT > & _eval, eoGenOp < MOEOT > & _op, moeoBinaryIndicatorBasedFitnessAssignment < MOEOT >& _fitnessAssignment) :
defaultGenContinuator(0), continuator(_continuator), eval(_eval), defaultPopEval(_eval), popEval(defaultPopEval), select(2),
selectMany(select,0.0), selectTransform(defaultSelect, defaultTransform), defaultSGAGenOp(defaultQuadOp, 1.0, defaultMonOp, 1.0), genBreed(select, _op), breed(genBreed), default_fitnessAssignment(NULL), fitnessAssignment(_fitnessAssignment), replace(fitnessAssignment, diversityAssignment)
{}
@ -260,8 +260,8 @@ protected:
/** breeder */
eoBreed < MOEOT > & breed;
/** fitness assignment used in IBEA */
moeoExpBinaryIndicatorBasedFitnessAssignment < MOEOT >& fitnessAssignment;
moeoExpBinaryIndicatorBasedFitnessAssignment < MOEOT >* default_fitnessAssignment;
moeoBinaryIndicatorBasedFitnessAssignment < MOEOT >& fitnessAssignment;
/** dummy diversity assignment */
moeoDummyDiversityAssignment < MOEOT > diversityAssignment;
/** environmental replacement */

121
moeo/src/continue/moeoDualHypContinue.h Normal file
View file

@ -0,0 +1,121 @@
/*
(c) 2013 Thales group
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; version 2
of the License.
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: http://eodev.sourceforge.net
Authors:
Johann Dréo <johann.dreo@thalesgroup.com>
*/
#ifndef _moeoDualHypContinue_h
#define _moeoDualHypContinue_h
#include <continue/moeoHypContinue.h>
/**
Continues until the (feasible or unfeasible) given Pareto set is reached.
@ingroup Continuators
*/
template< class MOEOT, class MetricT = moeoDualHyperVolumeDifferenceMetric<typename MOEOT::ObjectiveVector> >
class moeoDualHypContinue: public moeoHypContinue<MOEOT, MetricT >
{
protected:
bool is_feasible;
using moeoHypContinue<MOEOT, MetricT>::arch;
using moeoHypContinue<MOEOT, MetricT>::OptimSet;
using moeoHypContinue<MOEOT, MetricT>::pareto;
using moeoHypContinue<MOEOT, MetricT>::is_null_hypervolume;
public:
typedef typename MOEOT::ObjectiveVector ObjectiveVector;
typedef typename ObjectiveVector::Type AtomType;
/** A continuator that stops once a given Pareto front has been reached
*
* You should specify the feasibility of the targeted front.
* NOTE: the MOEOT::ObjectiveVector is supposed to implement the moeoDualRealObjectiveVector interface.
*
*/
moeoDualHypContinue( const std::vector<AtomType> & _OptimVec, bool _is_feasible, moeoArchive < MOEOT > & _archive, bool _normalize=true, double _rho=1.1 )
: moeoHypContinue<MOEOT, MetricT>( _OptimVec, _archive, _normalize, _rho ),
is_feasible(_is_feasible)
{
assert( _OptimVec.size() > 0);
vectorToParetoSet(_OptimVec);
}
/** A continuator that stops once a given Pareto front has been reached
*
* You should specify the feasibility of the targeted front.
* NOTE: the MOEOT::ObjectiveVector is supposed to implement the moeoDualRealObjectiveVector interface.
*
*/
moeoDualHypContinue( const std::vector<AtomType> & _OptimVec, bool _is_feasible, moeoArchive < MOEOT > & _archive, bool _normalize=true, ObjectiveVector& _ref_point=NULL )
: moeoHypContinue<MOEOT, MetricT>( _OptimVec, _archive, _normalize, _ref_point ),
is_feasible(_is_feasible)
{
assert( _OptimVec.size() > 0);
vectorToParetoSet(_OptimVec);
}
/** Returns false when a ParetoSet is reached. */
virtual bool operator() ( const eoPop<MOEOT>& /*_pop*/ )
{
std::vector<ObjectiveVector> bestCurrentParetoSet = pareto( arch );
#ifndef NDEBUG
assert( bestCurrentParetoSet.size() > 0 );
for( unsigned int i=1; i<bestCurrentParetoSet.size(); ++i ) {
assert( bestCurrentParetoSet[i].is_feasible() == bestCurrentParetoSet[0].is_feasible() );
}
#endif
// The current Pareto front is either feasible or unfeasible.
// It could not contains both kind of objective vectors, because a feasible solution always dominates an unfeasible front.
if( bestCurrentParetoSet[0].is_feasible() != OptimSet[0].is_feasible() ) {
return false;
}
return is_null_hypervolume( bestCurrentParetoSet );
}
protected:
/** Translate a vector given as param to the ParetoSet that should be reached. */
virtual void vectorToParetoSet(const std::vector<AtomType> & _OptimVec)
{
unsigned dim = (unsigned)(_OptimVec.size()/ObjectiveVector::Traits::nObjectives());
OptimSet.resize(dim);
unsigned k=0;
for(size_t i=0; i < dim; i++) {
for (size_t j=0; j < ObjectiveVector::Traits::nObjectives(); j++) {
// Use the feasibility declaration of an eoDualFitness
OptimSet[i][j] = AtomType(_OptimVec[k++], is_feasible);
}
}
}
};
#endif

113
moeo/src/continue/moeoHypContinue.h
View file

@ -36,7 +36,6 @@
//-----------------------------------------------------------------------------
#ifndef _moeoHypContinue_h
#define _moeoHypContinue_h
@ -60,20 +59,32 @@ public:
/// Ctor
moeoHypContinue( const std::vector<AtomType> & _OptimVec, moeoArchive < MOEOT > & _archive, bool _normalize=true, double _rho=1.1)
: eoContinue<MOEOT>(), arch(_archive), metric(_normalize,_rho)
: eoContinue<MOEOT>(), arch(_archive), default_metric(new MetricT(_normalize,_rho)), metric(*default_metric)
{
vectorToParetoSet(_OptimVec);
}
moeoHypContinue( const std::vector<AtomType> & _OptimVec, moeoArchive < MOEOT > & _archive, bool _normalize=true, ObjectiveVector& _ref_point=NULL)
: eoContinue<MOEOT> (), arch(_archive), metric(_normalize,_ref_point)
: eoContinue<MOEOT>(), arch(_archive), default_metric(new MetricT(_normalize,_ref_point)), metric(*default_metric)
{
vectorToParetoSet(_OptimVec);
}
moeoHypContinue( MetricT& _metric, const std::vector<AtomType> & _OptimVec, moeoArchive < MOEOT > & _archive )
: eoContinue<MOEOT>(), arch(_archive), default_metric(NULL), metric(_metric)
{
vectorToParetoSet(_OptimVec);
}
~moeoHypContinue()
{
if( default_metric != NULL ) {
delete default_metric;
}
}
/** Returns false when a ParetoSet is reached. */
virtual bool operator() ( const eoPop<MOEOT>& _pop )
virtual bool operator() ( const eoPop<MOEOT>& /*_pop*/ )
{
std::vector<ObjectiveVector> bestCurrentParetoSet = pareto( arch );
@ -88,8 +99,8 @@ protected:
{
std::vector < ObjectiveVector > bestCurrentParetoSet;
for (size_t i=0; i<arch.size(); i++) {
bestCurrentParetoSet.push_back(arch[i].objectiveVector());
for (size_t i=0; i<_archive.size(); i++) {
bestCurrentParetoSet.push_back(_archive[i].objectiveVector());
}
return bestCurrentParetoSet;
@ -123,96 +134,10 @@ protected:
protected:
moeoArchive <MOEOT> & arch;
MetricT metric;
MetricT* default_metric;
MetricT& metric;
std::vector <ObjectiveVector> OptimSet;
};
/**
Continues until the (feasible or unfeasible) given Pareto set is reached.
@ingroup Continuators
*/
template< class MOEOT, class MetricT = moeoDualHyperVolumeDifferenceMetric<typename MOEOT::ObjectiveVector> >
class moeoDualHypContinue: public moeoHypContinue<MOEOT, MetricT >
{
protected:
bool is_feasible;
using moeoHypContinue<MOEOT, MetricT>::arch;
using moeoHypContinue<MOEOT, MetricT>::OptimSet;
public:
typedef typename MOEOT::ObjectiveVector ObjectiveVector;
typedef typename ObjectiveVector::Type AtomType;
/** A continuator that stops once a given Pareto front has been reached
*
* You should specify the feasibility of the targeted front.
* NOTE: the MOEOT::ObjectiveVector is supposed to implement the moeoDualRealObjectiveVector interface.
*
*/
moeoDualHypContinue<MOEOT, MetricT>( const std::vector<AtomType> & _OptimVec, bool _is_feasible, moeoArchive < MOEOT > & _archive, bool _normalize=true, double _rho=1.1 )
: moeoHypContinue<MOEOT, MetricT>( _OptimVec, _archive, _normalize, _rho ), is_feasible(_is_feasible)
{
assert( _OptimVec.size() > 0);
vectorToParetoSet(_OptimVec);
}
/** A continuator that stops once a given Pareto front has been reached
*
* You should specify the feasibility of the targeted front.
* NOTE: the MOEOT::ObjectiveVector is supposed to implement the moeoDualRealObjectiveVector interface.
*
*/
moeoDualHypContinue<MOEOT, MetricT>( const std::vector<AtomType> & _OptimVec, bool _is_feasible, moeoArchive < MOEOT > & _archive, bool _normalize=true, ObjectiveVector& _ref_point=NULL )
: moeoHypContinue<MOEOT, MetricT>( _OptimVec, _archive, _normalize, _ref_point ), is_feasible(_is_feasible)
{
assert( _OptimVec.size() > 0);
vectorToParetoSet(_OptimVec);
}
/** Returns false when a ParetoSet is reached. */
virtual bool operator() ( const eoPop<MOEOT>& _pop )
{
std::vector<ObjectiveVector> bestCurrentParetoSet = pareto( arch );
#ifndef NDEBUG
assert( bestCurrentParetoSet.size() > 0 );
for( unsigned int i=1; i<bestCurrentParetoSet.size(); ++i ) {
assert( bestCurrentParetoSet[i].is_feasible() == bestCurrentParetoSet[0].is_feasible() );
}
#endif
// The current Pareto front is either feasible or unfeasible.
// It could not contains both kind of objective vectors, because a feasible solution always dominates an unfeasible front.
if( bestCurrentParetoSet[0].is_feasible() != OptimSet[0].is_feasible() ) {
return false;
}
return is_null_hypervolume( bestCurrentParetoSet );
}
protected:
using moeoHypContinue<MOEOT, MetricT>::pareto;
using moeoHypContinue<MOEOT, MetricT>::is_null_hypervolume;
/** Translate a vector given as param to the ParetoSet that should be reached. */
virtual void vectorToParetoSet(const std::vector<AtomType> & _OptimVec)
{
unsigned dim = (unsigned)(_OptimVec.size()/ObjectiveVector::Traits::nObjectives());
OptimSet.resize(dim);
unsigned k=0;
for(size_t i=0; i < dim; i++) {
for (size_t j=0; j < ObjectiveVector::Traits::nObjectives(); j++) {
// Use the feasibility declaration of an eoDualFitness
OptimSet[i][j] = AtomType(_OptimVec[k++], is_feasible);
}
}
}
};
#endif

2
moeo/src/diversity/moeoDummyDiversityAssignment.h
View file

@ -74,7 +74,7 @@ class moeoDummyDiversityAssignment : public moeoDiversityAssignment < MOEOT >
* @param _pop the population
* @param _objVec the objective vector
*/
void updateByDeleting(eoPop < MOEOT > & _pop, ObjectiveVector & _objVec)
void updateByDeleting(eoPop < MOEOT > & /*_pop*/, ObjectiveVector & /*_objVec*/)
{
// nothing to do... ;-)
}

109
moeo/src/fitness/moeoExpBinaryIndicatorBasedDualFitnessAssignment.h
View file

@ -1,3 +1,31 @@
/*
(c) 2013 Thales group
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; version 2
of the License.
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: http://eodev.sourceforge.net
Authors:
Johann Dréo <johann.dreo@thalesgroup.com>
*/
#ifndef MOEOEXPBINARYINDICATORBASEDDUALFITNESSASSIGNMENT_H_
#define MOEOEXPBINARYINDICATORBASEDDUALFITNESSASSIGNMENT_H_
#include <fitness/moeoExpBinaryIndicatorBasedFitnessAssignment.h>
@ -5,8 +33,7 @@ template<class MOEOT>
class moeoExpBinaryIndicatorBasedDualFitnessAssignment : public moeoExpBinaryIndicatorBasedFitnessAssignment<MOEOT>
{
protected:
eoPop<MOEOT> _feasible_pop;
eoPop<MOEOT> _unfeasible_pop;
eoDualPopSplit<MOEOT> _pop_split;
public:
typedef typename MOEOT::ObjectiveVector ObjectiveVector;
@ -19,21 +46,6 @@ public:
const double kappa = 0.05
) : moeoExpBinaryIndicatorBasedFitnessAssignment<MOEOT>( metric, kappa ) {}
//! Split up the population in two: in one pop the feasible individual, in the other the feasible ones
virtual void split( eoPop<MOEOT> & pop )
{
_feasible_pop.reserve(pop.size());
_unfeasible_pop.reserve(pop.size());
for( typename eoPop<MOEOT>::iterator it=pop.begin(), end=pop.end(); it != end; ++it ) {
// The ObjectiveVector should implement "is_feasible"
if( it->objectiveVector().is_feasible() ) {
_feasible_pop.push_back( *it );
} else {
_unfeasible_pop.push_back( *it );
}
}
}
/*! If the population is homogeneous (only composed of feasible individuals or unfeasible ones),
* then apply the operators on the whole population.
@ -41,42 +53,51 @@ public:
*/
virtual void operator()( eoPop<MOEOT>& pop )
{
// separate the pop in the members
split( pop );
// separate the pop in feasible/unfeasible
_pop_split( pop );
eoPop<MOEOT>* ppop;
// if there is at least one feasible individual, it will supersede all the unfeasible ones
if( _feasible_pop.size() == 0 ) {
ppop = & _unfeasible_pop;
// if there is at least one feasible individual,
// it will supersede all the unfeasible ones
if( _pop_split.feasible().size() == 0 ) {
ppop = & _pop_split.unfeasible();
} else {
ppop = & _feasible_pop;
ppop = & _pop_split.feasible();
}
this->setup(*ppop);
this->computeValues(*ppop);
this->setFitnesses(*ppop);
this->setFitnesses(*ppop); // NOTE: this alter individuals
// bring back altered individuals in the pop
pop = _pop_split.merge();
}
protected:
using moeoExpBinaryIndicatorBasedFitnessAssignment<MOEOT>::kappa;
/**
* Compute every indicator value in values (values[i] = I(_v[i], _o))
* @param _pop the population
*/
void computeValues(const eoPop < MOEOT > & _pop)
virtual void computeValues(const eoPop < MOEOT > & pop)
{
values.clear();
values.resize(_pop.size());
for (unsigned int i=0; i<_pop.size(); i++)
{
values[i].resize(_pop.size());
values.resize(pop.size());
for (unsigned int i=0; i<pop.size(); i++) {
values[i].resize(pop.size());
// the metric may not be symetric, thus neither is the matrix
for (unsigned int j=0; j<_pop.size(); j++)
{
if (i != j)
{
values[i][j] = Type( metric(_pop[i].objectiveVector(), _pop[j].objectiveVector()), _pop[i].objectiveVector().is_feasible() );
}
}
}
for (unsigned int j=0; j<pop.size(); j++) {
if (i != j) {
values[i][j] = Type(
metric( pop[i].objectiveVector(), pop[j].objectiveVector() ),
pop[i].objectiveVector().is_feasible()
);
} // if i != j
} // for j in pop
} // for i in pop
}
virtual void setFitnesses(eoPop < MOEOT > & pop)
@ -87,6 +108,20 @@ public:
}
}
virtual Type computeFitness(const unsigned int _idx)
{
Type result( 0.0, values[_idx][_idx].is_feasible() );
for (unsigned int i=0; i<values.size(); i++)
{
if (i != _idx)
{
result -= exp(-values[i][_idx]/kappa);
}
}
return result;
}
};
#endif // MOEOEXPBINARYINDICATORBASEDDUALFITNESSASSIGNMENT_H_

4
moeo/src/fitness/moeoExpBinaryIndicatorBasedFitnessAssignment.h
View file

@ -175,7 +175,7 @@ class moeoExpBinaryIndicatorBasedFitnessAssignment : public moeoBinaryIndicatorB
* Compute every indicator value in values (values[i] = I(_v[i], _o))
* @param _pop the population
*/
void computeValues(const eoPop < MOEOT > & _pop)
virtual void computeValues(const eoPop < MOEOT > & _pop)
{
values.clear();
values.resize(_pop.size());
@ -211,7 +211,7 @@ class moeoExpBinaryIndicatorBasedFitnessAssignment : public moeoBinaryIndicatorB
* Returns the fitness value of the _idx th individual of the population
* @param _idx the index
*/
Type computeFitness(const unsigned int _idx)
virtual Type computeFitness(const unsigned int _idx)
{
Type result(0.0);
for (unsigned int i=0; i<values.size(); i++)

117
moeo/src/metric/moeoDualHyperVolumeDifferenceMetric.h Normal file
View file

@ -0,0 +1,117 @@
/*
(c) 2013 Thales group
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; version 2
of the License.
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: http://eodev.sourceforge.net
Authors:
Johann Dréo <johann.dreo@thalesgroup.com>
*/
#ifndef MOEODUALHYPERVOLUMEDIFFERENCEMETRIC_H_
#define MOEODUALHYPERVOLUMEDIFFERENCEMETRIC_H_
#include <metric/moeoHyperVolumeDifferenceMetric.h>
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 /*MOEODUALHYPERVOLUMEDIFFERENCEMETRIC_H_*/

88
moeo/src/metric/moeoHyperVolumeDifferenceMetric.h
View file

@ -84,7 +84,7 @@ class moeoHyperVolumeDifferenceMetric : public moeoVectorVsVectorBinaryMetric <
* @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
*/
double operator()(const std::vector < ObjectiveVector > & _set1, const std::vector < ObjectiveVector > & _set2)
virtual double operator()(const std::vector < ObjectiveVector > & _set1, const std::vector < ObjectiveVector > & _set2)
{
double hypervolume_set1;
@ -197,90 +197,4 @@ class moeoHyperVolumeDifferenceMetric : public moeoVectorVsVectorBinaryMetric <
};
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
*/
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_*/

2
moeo/src/moeo
View file

@ -144,6 +144,7 @@
#include <metric/moeoEntropyMetric.h>
#include <metric/moeoHypervolumeBinaryMetric.h>
#include <metric/moeoHyperVolumeDifferenceMetric.h>
#include <metric/moeoDualHyperVolumeDifferenceMetric.h>
#include <metric/moeoHyperVolumeMetric.h>
#include <metric/moeoMetric.h>
#include <metric/moeoNormalizedSolutionVsSolutionBinaryMetric.h>
@ -217,5 +218,6 @@
#include <utils/moeoObjVecStat.h>
#include <continue/moeoHypContinue.h>
#include <continue/moeoDualHypContinue.h>
#endif /*MOEO_*/

35
moeo/src/utils/moeoBinaryMetricStat.h
View file

@ -1,5 +1,38 @@
/*
(c) 2013 Thales group
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; version 2
of the License.
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: http://eodev.sourceforge.net
Authors:
Johann Dréo <johann.dreo@thalesgroup.com>
*/
#ifndef _MOEOBINARYMETRICSTAT_H_
#define _MOEOBINARYMETRICSTAT_H_
#include <eo>
/** A wrapper to save a moeoMetric in an eoStat
*
* This wrap a MOEO binary metric into an eoStat
* This is useful if you want to use it in a checkpoint, for instance.
*/
template <class MOEOT, class T = double>
class moeoBinaryMetricStat : public eoStat<MOEOT, T>
{
@ -57,3 +90,5 @@ protected:
bool _first_gen;
};
#endif // _MOEOBINARYMETRICSTAT_H_