STK::IMixtureStatModel Class Reference

Interface base class for Mixture (composed) model. More...

#include <STK_IMixtureStatModel.h>

Inheritance diagram for STK::IMixtureStatModel:

Public Types
typedef std::vector< IMixture * >::const_iterator	ConstMixtIterator
typedef std::vector< IMixture * >::iterator	MixtIterator

Public Member Functions
virtual	~IMixtureStatModel ()
	destructor
int	nbCluster () const
CPointX const &	pk () const
CArrayXX const &	tik () const
CPointX const &	tk () const
CVectorXi const &	zi () const
std::vector< IMixture * > const &	v_mixtures () const
Real	computeLnLikelihood (int i) const
Real	computeLikelihood (int i) const
Real	computeLnLikelihood () const
Real	computeICL () const
IMixture *	getMixture (String const &idData) const
	Utility lookup function allowing to find a Mixture from its idData.
void	registerMixture (IMixture *p_mixture)
	register a mixture to the composer.
void	releaseMixture (String const &idData)
	release a mixture from the composer.
int	computeNbFreeParameters () const
	compute the number of free parameters of the model.
int	computeNbMissingValues () const
	compute the missing values of the model.
virtual IMixtureStatModel *	create () const =0
	create pattern
virtual IMixtureStatModel *	clone () const =0
	clone pattern
virtual void	randomInit ()=0
	initialize randomly the parameters of the components of the model
virtual Real	lnComponentProbability (int i, int k) const =0
virtual void	initializeStep ()
	Initialize the model before at its first use.
virtual void	imputationStep ()
	Impute the missing values.
virtual void	samplingStep ()
	Simulation of all the latent variables and/or missing data excluding class labels.
virtual void	setParametersStep ()
	Utility method allowing to signal to a mixture to set its parameters.
virtual void	storeIntermediateResults (int iteration)
	This step can be used to signal to the mixtures that they must store results.
virtual void	releaseIntermediateResults ()
	This step can be used to signal to the mixtures that they must release the stored results.
virtual void	finalizeStep ()
	Finalize the estimation of the model.
virtual void	writeParameters (ostream &os) const
	write the parameters of the model in the stream os.
template<class Array >
void	setMixtureParameters (Array const &tik)
	set the mixture parameters using an array of posterior probabilities.
template<class Array , class RowVector >
void	setMixtureParameters (Array const &tik, RowVector const &pk)
	set the mixture parameters giving the posterior probabilities and the proportions.
template<class RowVector >
void	setProportions (RowVector const &pk)
	Set proportions of each classes.
template<class Manager , class Parameters >
void	setParameters (IMixtureManager< Manager > const &manager, String const &idData, Parameters const &param)
	Utility method allowing to set the parameters to a specific mixture.
template<class Manager >
void	createMixture (IMixtureManager< Manager > &manager)
	Utility method allowing to create all the mixtures registered in the data handler of a mixture manager and to register them.
template<class Manager >
IMixture *	createMixture (IMixtureManager< Manager > &manager, String const &idData)
	Utility method allowing to create a mixture with a given data set and register it.
template<class Manager >
void	removeMixture (IMixtureManager< Manager > &manager, String const &idData)
	Utility method allowing to release completely a mixture with its data set.
template<class Manager , class Parameters >
void	getParameters (IMixtureManager< Manager > const &manager, String const &idData, Parameters &param) const
	Utility method allowing to get the parameters of a specific mixture.
template<class Manager , class MissingValues >
void	getMissingValues (IMixtureManager< Manager > const &manager, String const &idData, MissingValues &missing) const
	Utility method allowing to get the missing values of a specific mixture.
template<class DataHandler >
void	createMixture (IMixtureManager< DataHandler > &manager)
Public Member Functions inherited from STK::IStatModelBase
int	nbSample () const
Real	lnNbSample () const
int	nbVariable () const
Real	lnLikelihood () const
Real	likelihood () const
int	nbFreeParameter () const
Real	computeBIC () const
Real	computeAIC () const
Real	computeML () const

Protected Member Functions
	IMixtureStatModel (int nbSample, int nbCluster)
	Constructor.
	IMixtureStatModel (IMixtureStatModel const &model)
	copy constructor.
void	setNbCluster (int nbCluster)
	set the number of cluster of the model
Protected Member Functions inherited from STK::IStatModelBase
	IStatModelBase ()
	Default constructor.
	IStatModelBase (int nbSample)
	Constructor with specified dimension.
	IStatModelBase (int nbSample, int nbVariable)
	Constructor with specified dimension.
	IStatModelBase (IStatModelBase const &model)
	Copy constructor.
	~IStatModelBase ()
	destructor
void	setNbFreeParameter (int const &nbFreeParameter)
	set the number of free parameters of the model
void	setNbSample (int const &nbSample)
	set the number of samples of the model
void	setNbVariable (int const &nbVariable)
	set the number of variables of the model
void	setLnLikelihood (Real const &lnLikelihood)
	set the log-likelihood of the model
void	initialize (int nbSample, int nbVariable)
	set the dimensions of the parameters of the model

Protected Attributes
int	nbCluster_
	number of cluster.
CPointX	pk_
	The proportions of each mixtures.
CArrayXX	tik_
	The tik probabilities.
CPointX	tk_
	The sum of the columns of tik_.
CVectorXi	zi_
	The zi class label.
std::vector< IMixture * >	v_mixtures_
	vector of pointers to the mixtures components

Detailed Description

Interface base class for Mixture (composed) model.

In statistics, a mixture model is a probabilistic model for representing the presence of sub-populations within an overall population, without requiring that an observed data-set should identify the sub-population to which an individual observation belongs. Formally a mixture model corresponds to the mixture distribution that represents the probability distribution of observations in the overall population. However, while problems associated with "mixture distributions" relate to deriving the properties of the overall population from those of the sub-populations, "mixture models" are used to make statistical inferences about the properties of the sub-populations given only observations on the pooled population, without sub-population-identity information.

Some ways of implementing mixture models involve steps that attribute postulated sub-population-identities to individual observations (or weights towards such sub-populations), in which case these can be regarded as types unsupervised learning or clustering procedures. However not all inference procedures involve such steps.

Pure virtual functions to implement in derived class are

virtual IMixtureComposer* create() const = 0;
virtual IMixtureComposer* clone() const = 0;
virtual bool randomInit() =0;
virtual Real lnComponentProbability(int i, int k) const = 0;

Virtual functions that can be re-implemented in derived class for a specific behavior are:

virtual void imputationStep();
virtual void samplingStep();
virtual void setParametersStep();
virtual void storeIntermediateResults(int iteration);
virtual void releaseIntermediateResults()
virtual void writeParameters(std::ostream& os) const;

Template functions allowing to interact with the composer are

template<class Array>
void setMixtureParameters( Array const& tik);
template<class Array, class RowVector>
void setMixtureParameters( Array const& tik, RowVector const& pk);
template<class RowVector>
void setProportions( RowVector const& pk);
 
template<class Manager, class Parameters>
void setParameters(IMixtureManager<Manager> const& manager, String const& idData, Parameters const& param);
 
template<class Manager>
void createMixture(IMixtureManager<Manager>& manager);
template<class DataHandler>
IMixture* createMixture(IMixtureManager<Manager>& manager, String const& idData);
template<class DataHandler>
void removeMixture(IMixtureManager<Manager>& manager, String const& idData);

Template functions allowing to get results from the composer are

template<class Manager, class Parameters>
void getParameters(IMixtureManager<Manager> const& manager, String const& idData, Parameters& param) const;
template<class Manager, class MissingValues>
void getMissingValues(IMixtureManager<Manager> const& manager, String const& idData, MissingValues& missing) const;

See also

IMixtureComposer, IMixtureLearner, IMixtureManager

Note

the virtual method IMixtureStatModel::initializeStep is called in all the initialization method. Don't forget to called it in the randomInit implementation.

Definition at line 123 of file STK_IMixtureStatModel.h.

Member Typedef Documentation

ConstMixtIterator

typedef std::vector<IMixture*>::const_iterator STK::IMixtureStatModel::ConstMixtIterator

Definition at line 136 of file STK_IMixtureStatModel.h.

MixtIterator

typedef std::vector<IMixture*>::iterator STK::IMixtureStatModel::MixtIterator

Definition at line 137 of file STK_IMixtureStatModel.h.

Constructor & Destructor Documentation

IMixtureStatModel() [1/2]

STK::IMixtureStatModel::IMixtureStatModel ( int  nbSample, int  nbCluster  )

protected

Constructor.

Parameters

nbCluster,nbSample

number of clusters and samples

Definition at line 45 of file STK_IMixtureStatModel.cpp.

                                    : IStatModelBase(nbSample)
                                    , nbCluster_(nbCluster)
                                    , pk_(nbCluster, 1./nbCluster)
                                    , tik_(nbSample, nbCluster, 1./nbCluster)
                                    , tk_(nbCluster, Real(nbSample)/nbCluster)
                                    , zi_(nbSample, baseIdx)
                                    , v_mixtures_()
{}

IMixtureStatModel() [2/2]

STK::IMixtureStatModel::IMixtureStatModel ( IMixtureStatModel const &  model )

protected

copy constructor.

Parameters

model

the model to clone

Definition at line 56 of file STK_IMixtureStatModel.cpp.

                                    : IStatModelBase(model)
                                    , nbCluster_(model.nbCluster_)
                                    , pk_(model.pk_), tik_(model.tik_)
                                    , tk_(model.tk_), zi_(model.zi_)
                                    , v_mixtures_(model.v_mixtures_.size())
{
  // clone mixtures
  for (size_t l = 0; l < v_mixtures_.size(); ++l)
  {
    v_mixtures_[l] = model.v_mixtures_[l]->clone();
    v_mixtures_[l]->setMixtureModel(this);
  }
}

References v_mixtures_.

~IMixtureStatModel()

STK::IMixtureStatModel::~IMixtureStatModel ( )

virtual

destructor

Definition at line 71 of file STK_IMixtureStatModel.cpp.

{
  for (MixtIterator it = v_mixtures_.begin() ; it != v_mixtures_.end(); ++it)
  { delete (*it);}
}

References v_mixtures_.

Member Function Documentation

clone()

virtual IMixtureStatModel * STK::IMixtureStatModel::clone ( ) const

pure virtual

clone pattern

Implemented in STK::MixtureComposer, STK::MixtureComposerFixedProp, STK::MixtureLearner, STK::MixtureSemiLearner, STK::MixtureSemiLearnerFixedProp, and STK::IMixtureComposer.

computeICL()

Real STK::IMixtureStatModel::computeICL

(

)

const

Returns

the computed ICL criteria.

Definition at line 104 of file STK_IMixtureStatModel.cpp.

{
  Real res = 0.0;
  for (int j = tik().beginCols(); j< tik().endCols(); ++j)
  { res += (tik_.col(j) * (tik_.col(j)+1e-15).log()).sum();}
  // compute result
  return (- 2. * lnLikelihood() + nbFreeParameter() * lnNbSample() - 2. * res);
}

References STK::ICArray< Derived >::col(), STK::IStatModelBase::lnLikelihood(), STK::IStatModelBase::lnNbSample(), STK::IStatModelBase::nbFreeParameter(), STK::sum(), tik(), and tik_.

Referenced by STK::ICLMixtureCriterion::run().

computeLikelihood()

Real STK::IMixtureStatModel::computeLikelihood

(

int

i

)

const

Returns

the computed likelihood of the i-th sample.

Parameters

i	index of the sample

Definition at line 77 of file STK_IMixtureStatModel.cpp.

{ return std::exp(computeLnLikelihood(i));}

References computeLnLikelihood().

computeLnLikelihood() [1/2]

Real STK::IMixtureStatModel::computeLnLikelihood

(

)

const

Returns

the computed log-likelihood.

Definition at line 95 of file STK_IMixtureStatModel.cpp.

{
  Real res = 0.0;
  for (int i = tik().beginRows(); i< tik().endRows(); ++i)
  { res += computeLnLikelihood(i);}
  return res;
}

References computeLnLikelihood(), and tik().

Referenced by computeLikelihood(), computeLnLikelihood(), STK::IMixtureComposer::finalizeStep(), finalizeStep(), STK::IMixtureComposer::initializeStep(), STK::MixtureLearner::paramUpdateStep(), and releaseMixture().

computeLnLikelihood() [2/2]

Real STK::IMixtureStatModel::computeLnLikelihood

(

int

i

)

const

Returns

the computed log-likelihood of the i-th sample.

Parameters

i	index of the sample

Definition at line 83 of file STK_IMixtureStatModel.cpp.

{
  // get maximal value
  CPointX lnComp(pk_.size());
  for (int k = pk_.begin(); k< pk_.end(); ++k)
  { lnComp[k] = std::log(pk_[k]) + lnComponentProbability(i, k);}
  // compute result
  Real lnCompMax = lnComp.maxElt();
  return std::log((lnComp-lnCompMax).exp().sum())+lnCompMax;
}

References lnComponentProbability(), pk_, and STK::sum().

computeNbFreeParameters()

int STK::IMixtureStatModel::computeNbFreeParameters

(

)

const

compute the number of free parameters of the model.

This method is used in IMixtureComposer::initializeStep in order to give a value to IStatModelBase::nbFreeParameter_. lookup on the mixtures and sum the nbFreeParameter.

Returns

the number of free parameters

Definition at line 162 of file STK_IMixtureStatModel.cpp.

{
  int sum = nbCluster_-1; // proportions
  for (ConstMixtIterator it = v_mixtures_.begin(); it != v_mixtures_.end(); ++it)
  { sum+= (*it)->nbFreeParameter();}
  return sum;
}

References nbCluster_, STK::sum(), and v_mixtures_.

Referenced by STK::MixtureLearner::createLearner(), and STK::IMixtureComposer::initializeStep().

computeNbMissingValues()

int STK::IMixtureStatModel::computeNbMissingValues

(

)

const

compute the missing values of the model.

lookup on the mixtures and sum the nbMissingValues.

Returns

the number of missing values

Definition at line 174 of file STK_IMixtureStatModel.cpp.

{
  int sum = nbCluster_-1; // proportions
  for (ConstMixtIterator it = v_mixtures_.begin(); it != v_mixtures_.end(); ++it)
  { sum+= (*it)->nbMissingValues();}
  return sum;
}

References nbCluster_, STK::sum(), and v_mixtures_.

Referenced by STK::EMPredict::run(), STK::SemiSEMPredict::run(), and STK::MixtureComposer::writeParameters().

create()

virtual IMixtureStatModel * STK::IMixtureStatModel::create ( ) const

pure virtual

create pattern

Implemented in STK::MixtureComposer, STK::MixtureComposerFixedProp, STK::MixtureLearner, STK::MixtureSemiLearner, STK::MixtureSemiLearnerFixedProp, and STK::IMixtureComposer.

createMixture() [1/3]

template<class DataHandler >

void STK::IMixtureStatModel::createMixture

(

IMixtureManager< DataHandler > &

manager

)

Definition at line 348 of file STK_IMixtureStatModel.h.

{
  typedef typename DataHandlerBase<DataHandler>::InfoMap InfoMap;
  typedef typename InfoMap::const_iterator const_iterator;
  for ( const_iterator it=manager.p_handler()->info().begin(); it!=manager.p_handler()->info().end(); ++it)
  {
    IMixture* p_mixture = manager.createMixture(it->first, nbCluster());
#ifdef STK_MIXTURE_DEBUG
  if (!p_mixture)
  { stk_cout << _T("In IMixtureStatModel::createMixture(manager) failed.\n");}
#endif
    if (p_mixture) registerMixture(p_mixture);
  }
}

References _T, nbCluster(), registerMixture(), and stk_cout.

createMixture() [2/3]

template<class Manager >

void STK::IMixtureStatModel::createMixture

(

IMixtureManager< Manager > &

manager

)

Utility method allowing to create all the mixtures registered in the data handler of a mixture manager and to register them.

Parameters

manager

the manager with the responsibility of the creation.

createMixture() [3/3]

template<class Manager >

IMixture * STK::IMixtureStatModel::createMixture	(	IMixtureManager< Manager > &	manager,
		String const &	idData
	)

Utility method allowing to create a mixture with a given data set and register it.

The Mixture Manager will find the associated model to use with this data set.

Parameters

manager	the manager with the responsibility of the creation.
idData	the id name of the data to modelize.

Definition at line 370 of file STK_IMixtureStatModel.h.

{
  IMixture* p_mixture = manager.createMixture( idData, nbCluster());
#ifdef STK_MIXTURE_DEBUG
  if (!p_mixture)
  { stk_cout << _T("In IMixtureStatModel::createMixture(manager,")<< idData << _T(") failed.\n");}
#endif
  if (p_mixture) registerMixture(p_mixture);
  return p_mixture;
}

References _T, nbCluster(), registerMixture(), and stk_cout.

finalizeStep()

virtual void STK::IMixtureStatModel::finalizeStep ( )

inlinevirtual

Finalize the estimation of the model.

The default behavior is compute current lnLikelihood.

Reimplemented in STK::IMixtureComposer, STK::MixtureComposer, and STK::MixtureLearner.

Definition at line 247 of file STK_IMixtureStatModel.h.

{ setLnLikelihood(computeLnLikelihood());}

References computeLnLikelihood(), and STK::IStatModelBase::setLnLikelihood().

Referenced by STK::ImputeAlgo::run(), STK::SimulAlgo::run(), and STK::LearnFacade::run().

getMissingValues()

template<class Manager , class MissingValues >

void STK::IMixtureStatModel::getMissingValues	(	IMixtureManager< Manager > const &	manager,
		String const &	idData,
		MissingValues &	missing
	)		const

Utility method allowing to get the missing values of a specific mixture.

Parameters

manager	the manager with the responsibility of the parameters
idData	the Id of the data we want the parameters
missing	the structure which will receive the missing values

Definition at line 474 of file STK_IMixtureStatModel.h.

{
  IMixture* p_mixture= getMixture(idData);
  if (p_mixture) manager.getMissingValues(p_mixture, missing);
}

References getMixture().

getMixture()

IMixture * STK::IMixtureStatModel::getMixture

(

String const &

idData

)

const

Utility lookup function allowing to find a Mixture from its idData.

Parameters

idData

the id name of the data

Returns

a pointer on the mixture, NULL if the mixture is not found

Definition at line 117 of file STK_IMixtureStatModel.cpp.

{
  for (ConstMixtIterator it = v_mixtures_.begin(); it != v_mixtures_.end(); ++it)
  { if ((*it)->idData() == idData) return (*it);}
  return 0;
}

References v_mixtures_.

Referenced by getMissingValues(), getParameters(), removeMixture(), and setParameters().

getParameters()

template<class Manager , class Parameters >

void STK::IMixtureStatModel::getParameters	(	IMixtureManager< Manager > const &	manager,
		String const &	idData,
		Parameters &	param
	)		const

Utility method allowing to get the parameters of a specific mixture.

Parameters

manager	the manager with the responsibility of the parameters
idData	the Id of the data we want the parameters
param	the structure which will receive the parameters

Definition at line 464 of file STK_IMixtureStatModel.h.

{  IMixture* p_mixture= getMixture(idData);
   if (p_mixture) manager.getParameters(p_mixture, param);
}

References getMixture().

imputationStep()

virtual void STK::IMixtureStatModel::imputationStep ( )

inlinevirtual

Impute the missing values.

Default behavior is "do nothing".

Reimplemented in STK::MixtureComposer, and STK::MixtureLearner.

Definition at line 225 of file STK_IMixtureStatModel.h.

{}

Referenced by STK::EMAlgo::run(), STK::CEMAlgo::run(), STK::ImputeAlgo::run(), and STK::EMPredict::run().

initializeStep()

void STK::IMixtureStatModel::initializeStep ( )

virtual

Initialize the model before at its first use.

This function can be overloaded in derived class for initialization of the specific model parameters. It should be called prior to any used of the class.

See also

IMixture,MixtureBridge,MixtureLearner

Reimplemented in STK::IMixtureComposer.

Definition at line 188 of file STK_IMixtureStatModel.cpp.

{
#ifdef STK_MIXTURE_VERBOSE
  stk_cout << _T("Entering IMixtureStatModel::initializeStep\n");
#endif
  if (v_mixtures_.size() == 0)
    STKRUNTIME_ERROR_NO_ARG(IMixtureStatModel::initializeStep,no mixture registered);
  setLnLikelihood(-Arithmetic<Real>::infinity());
  // initialize registered mixtures
  for (MixtIterator it = v_mixtures_.begin(); it != v_mixtures_.end(); ++it)
  { (*it)->initializeStep();}
}

References _T, initializeStep(), STK::IStatModelBase::setLnLikelihood(), stk_cout, STKRUNTIME_ERROR_NO_ARG, and v_mixtures_.

Referenced by STK::IMixtureComposer::initializeStep(), initializeStep(), STK::MixtureLearner::randomInit(), and STK::LearnFacade::run().

lnComponentProbability()

virtual Real STK::IMixtureStatModel::lnComponentProbability ( int  i, int  k  ) const

pure virtual

Returns

the value of the probability of the ith sample in the kth class.

Parameters

i,k	indexes of the sample and of the class

Implemented in STK::MixtureComposer, and STK::MixtureLearner.

Referenced by computeLnLikelihood(), STK::IMixtureComposer::eStep(), and STK::IMixtureLearner::mapStep().

nbCluster()

int STK::IMixtureStatModel::nbCluster ( ) const

inline

Returns

the number of cluster

Definition at line 143 of file STK_IMixtureStatModel.h.

{ return nbCluster_;}

References nbCluster_.

Referenced by STK::MixtureComposer::create(), STK::MixtureComposerFixedProp::create(), STK::MixtureLearner::create(), STK::MixtureSemiLearner::create(), STK::MixtureSemiLearnerFixedProp::create(), createMixture(), createMixture(), STK::MixtureSemiLearner::initializeMixtureParameters(), STK::MixtureComposer::MixtureComposer(), STK::MixtureLearner::MixtureLearner(), setNbCluster(), setProportions(), STK::MixtureComposer::writeParameters(), and STK::MixtureLearner::writeParameters().

pk()

CPointX const & STK::IMixtureStatModel::pk ( ) const

inline

Returns

the proportions of each mixtures

Definition at line 145 of file STK_IMixtureStatModel.h.

{ return pk_;};

References pk_.

Referenced by STK::IMixtureLearner::setClassLabels(), setMixtureParameters(), setProportions(), STK::MixtureComposer::writeParameters(), and STK::MixtureLearner::writeParameters().

randomInit()

virtual void STK::IMixtureStatModel::randomInit ( )

pure virtual

initialize randomly the parameters of the components of the model

Implemented in STK::MixtureComposer, STK::MixtureLearner, and STK::IMixtureComposer.

registerMixture()

void STK::IMixtureStatModel::registerMixture

(

IMixture *

p_mixture

)

register a mixture to the composer.

When a mixture is registered, the composer:

• assign composer pointer (itself) to the mixture
• add it to v_mixtures_
• update the number of variables

  Parameters

  p_mixture

  a pointer on the mixture

Definition at line 124 of file STK_IMixtureStatModel.cpp.

{
#ifdef STK_MIXTURE_VERBOSE
  stk_cout << _T("In IMixtureStatModel::registerMixture, registering mixture: ")
           << p_mixture->idData() << _T("\n");
#endif
  p_mixture->setMixtureModel(this);
  v_mixtures_.push_back(p_mixture);
  // update NbFreeParameters
  setNbFreeParameter(nbFreeParameter()+p_mixture->nbFreeParameter());
}

References _T, STK::IRegression< YArray, XArray, Weights >::nbFreeParameter(), STK::IStatModelBase::nbFreeParameter(), STK::IStatModelBase::setNbFreeParameter(), stk_cout, and v_mixtures_.

Referenced by createMixture(), and createMixture().

releaseIntermediateResults()

virtual void STK::IMixtureStatModel::releaseIntermediateResults ( )

inlinevirtual

This step can be used to signal to the mixtures that they must release the stored results.

This is usually called if the estimation process failed.

Reimplemented in STK::MixtureComposer, and STK::MixtureLearner.

Definition at line 243 of file STK_IMixtureStatModel.h.

{}

Referenced by STK::SEMAlgo::run(), STK::SemiSEMAlgo::run(), and STK::SimulAlgo::run().

releaseMixture()

void STK::IMixtureStatModel::releaseMixture

(

String const &

idData

)

release a mixture from the composer.

When a mixture is released, the composer remove it from v_mixtures_.

Note

the data set associated with the mixture is still in the manager list of data set if you use a manager in order to register it.

Parameters

idData

the Id of the mixture to release.

Definition at line 140 of file STK_IMixtureStatModel.cpp.

{
  for (MixtIterator it = v_mixtures_.begin(); it != v_mixtures_.end(); ++it)
  {
    if ((*it)->idData() == idData)
    {
       setNbFreeParameter(nbFreeParameter()-(*it)->nbFreeParameter());
       // remove mixture
       delete (*it);
       v_mixtures_.erase(it);
       // update log-likelihood
       if (v_mixtures_.size() == 0)
       { setLnLikelihood(-Arithmetic<Real>::infinity());}
       else
       { setLnLikelihood(computeLnLikelihood());}
       // and break
       break;
    }
  }
}

References computeLnLikelihood(), STK::IStatModelBase::nbFreeParameter(), STK::IStatModelBase::setLnLikelihood(), STK::IStatModelBase::setNbFreeParameter(), and v_mixtures_.

Referenced by removeMixture().

removeMixture()

template<class Manager >

void STK::IMixtureStatModel::removeMixture	(	IMixtureManager< Manager > &	manager,
		String const &	idData
	)

Utility method allowing to release completely a mixture with its data set.

The MixtureManager will find and release the associated data set.

Parameters

manager	the manager with the responsibility of the release.
idData	the id name of the data to modelize.

Definition at line 387 of file STK_IMixtureStatModel.h.

{
  IMixture* p_mixture = getMixture(idData);
#ifdef STK_MIXTURE_DEBUG
  if (!p_mixture)
  { stk_cout << _T("In IMixtureStatModel::removeMixture(manager,")<< idData << _T(") failed.\n");}
#endif
  if (p_mixture)
  {
    releaseMixture(idData);
    manager.releaseDataBridge( idData);
  }
}

References _T, getMixture(), releaseMixture(), and stk_cout.

samplingStep()

virtual void STK::IMixtureStatModel::samplingStep ( )

inlinevirtual

Simulation of all the latent variables and/or missing data excluding class labels.

Default behavior is "do nothing".

Reimplemented in STK::MixtureComposer, and STK::MixtureLearner.

Definition at line 229 of file STK_IMixtureStatModel.h.

{}

Referenced by STK::IMixtureAlgoPredict::burnStep(), STK::SEMAlgo::run(), STK::SemiSEMAlgo::run(), STK::SimulAlgo::run(), and STK::SemiSEMPredict::run().

setMixtureParameters() [1/2]

template<class Array >

void STK::IMixtureStatModel::setMixtureParameters

(

Array const &

tik

)

set the mixture parameters using an array of posterior probabilities.

Proportions, numbers in each class and class labels are computed using these posterior probabilities.

Parameters

tik	posterior class probabilities

Definition at line 414 of file STK_IMixtureStatModel.h.

{
  tik_ = tik;
  tk_ = Stat::sumByCol(tik_);
  pk_ = tk_ / nbSample();
  for (int i = tik_.beginCols(); i< tik_.endCols(); ++i)
  {
    int k;
    tik_.row(i).maxElt(k);
    zi_[i] = k;
  }
}

References STK::IStatModelBase::nbSample(), pk_, STK::ICArray< Derived >::row(), STK::Stat::sumByCol(), tik(), tik_, tk_, and zi_.

Referenced by setMixtureParameters().

setMixtureParameters() [2/2]

template<class Array , class RowVector >

void STK::IMixtureStatModel::setMixtureParameters	(	Array const &	tik,
		RowVector const &	pk
	)

set the mixture parameters giving the posterior probabilities and the proportions.

Numbers in each class and class labels are computed using the posterior probabilities.

Parameters

tik	posterior class probabilities
pk	prior class proportion

Definition at line 430 of file STK_IMixtureStatModel.h.

{
#ifdef STK_MIXTURE_DEBUG
  if (pk_.size() != tik_.sizeCols())
  { STKRUNTIME_ERROR_2ARG(IMixtureLearner::setMixtureParameters,pk_.size(),tik_.sizeCols(),Numbers of class in tik and pk differ);}
#endif
  tik_ = tik;
  pk_  = pk;
  tk_  = Stat::sumByCol(tik_);
  for (int i = tik_.beginCols(); i< tik_.endCols(); ++i)
  {
    int k;
    tik_.row(i).maxElt(k);
    zi_[i] = k;
  }
}

References pk(), pk_, STK::ICArray< Derived >::row(), setMixtureParameters(), STKRUNTIME_ERROR_2ARG, STK::Stat::sumByCol(), tik(), tik_, tk_, and zi_.

setNbCluster()

void STK::IMixtureStatModel::setNbCluster ( int  nbCluster )

inlineprotected

set the number of cluster of the model

Parameters

nbCluster

number of cluster of the model

Definition at line 326 of file STK_IMixtureStatModel.h.

{ nbCluster_ = nbCluster;}

References nbCluster(), and nbCluster_.

setParameters()

template<class Manager , class Parameters >

void STK::IMixtureStatModel::setParameters	(	IMixtureManager< Manager > const &	manager,
		String const &	idData,
		Parameters const &	param
	)

Utility method allowing to set the parameters to a specific mixture.

Parameters

manager	the manager with the responsibility of the parameters
idData	Id of the data we want to set the parameters
param	structure which contains the parameters

Definition at line 407 of file STK_IMixtureStatModel.h.

{ IMixture* p_mixture= getMixture(idData);
  if (p_mixture) manager.setParameters(p_mixture, param);
}

References getMixture().

setParametersStep()

virtual void STK::IMixtureStatModel::setParametersStep ( )

inlinevirtual

Utility method allowing to signal to a mixture to set its parameters.

It will be called once enough intermediate results have been stored.

Reimplemented in STK::MixtureComposer, and STK::MixtureLearner.

Definition at line 233 of file STK_IMixtureStatModel.h.

{}

Referenced by STK::IMixtureAlgoPredict::burnStep(), STK::SEMAlgo::run(), STK::SemiSEMAlgo::run(), and STK::SemiSEMPredict::run().

setProportions()

template<class RowVector >

void STK::IMixtureStatModel::setProportions

(

RowVector const &

pk

)

Set proportions of each classes.

Parameters

pk	prior class proportion

Definition at line 449 of file STK_IMixtureStatModel.h.

{
#ifdef STK_MIXTURE_DEBUG
  if (pk_.size() != nbCluster())
  { STKRUNTIME_ERROR_2ARG(IMixtureLearner::setProportions,pk_.size(),nbCluster(),Numbers of class in pk differs);}
#endif
  pk_  = pk;
}

References nbCluster(), pk(), pk_, setProportions(), and STKRUNTIME_ERROR_2ARG.

Referenced by setProportions().

storeIntermediateResults()

virtual void STK::IMixtureStatModel::storeIntermediateResults ( int  iteration )

inlinevirtual

This step can be used to signal to the mixtures that they must store results.

This is usually called after a burn-in phase. The composer store the current value of the log-Likelihood.

Reimplemented in STK::MixtureComposer, and STK::MixtureLearner.

Definition at line 238 of file STK_IMixtureStatModel.h.

{}

Referenced by STK::IMixtureAlgoPredict::burnStep(), STK::SEMAlgo::run(), STK::SemiSEMAlgo::run(), STK::SimulAlgo::run(), and STK::SemiSEMPredict::run().

tik()

CArrayXX const & STK::IMixtureStatModel::tik ( ) const

inline

Returns

the tik probabilities

Definition at line 147 of file STK_IMixtureStatModel.h.

{ return tik_;};

References tik_.

Referenced by computeICL(), computeLnLikelihood(), setMixtureParameters(), and setMixtureParameters().

tk()

CPointX const & STK::IMixtureStatModel::tk ( ) const

inline

Returns

the proportions of individuals

Definition at line 149 of file STK_IMixtureStatModel.h.

{ return tk_;};

References tk_.

v_mixtures()

std::vector< IMixture * > const & STK::IMixtureStatModel::v_mixtures ( ) const

inline

Returns

a constant reference on the vector of mixture

Definition at line 153 of file STK_IMixtureStatModel.h.

{ return v_mixtures_;}

References v_mixtures_.

Referenced by STK::MixtureSemiLearner::create(), STK::MixtureSemiLearnerFixedProp::create(), STK::MixtureComposer::createComposer(), and STK::MixtureLearner::createLearner().

writeParameters()

virtual void STK::IMixtureStatModel::writeParameters ( ostream &  os ) const

inlinevirtual

write the parameters of the model in the stream os.

Reimplemented in STK::MixtureComposer, and STK::MixtureLearner.

Definition at line 249 of file STK_IMixtureStatModel.h.

{};

Referenced by STK::IMixtureComposer::eStep(), STK::IMixtureComposer::initializeStep(), STK::SimulAlgo::run(), and STK::FullStrategy::run().

zi()

CVectorXi const & STK::IMixtureStatModel::zi ( ) const

inline

Returns

the zi class label

Definition at line 151 of file STK_IMixtureStatModel.h.

{ return zi_;};

References zi_.

Referenced by STK::IMixtureLearner::setClassLabels(), STK::IMixtureLearner::setClassLabels(), and STK::MixtureSemiLearner::setLabels().

Member Data Documentation

nbCluster_

int STK::IMixtureStatModel::nbCluster_

protected

number of cluster.

Definition at line 329 of file STK_IMixtureStatModel.h.

Referenced by computeNbFreeParameters(), computeNbMissingValues(), STK::IMixtureComposer::initializeMixtureParameters(), nbCluster(), and setNbCluster().

pk_

CPointX STK::IMixtureStatModel::pk_

protected

The proportions of each mixtures.

Definition at line 331 of file STK_IMixtureStatModel.h.

Referenced by computeLnLikelihood(), STK::MixtureSemiLearner::eStep(), STK::IMixtureComposer::eStep(), STK::IMixtureComposer::initializeMixtureParameters(), STK::IMixtureLearner::mapStep(), pk(), STK::IMixtureComposer::pStep(), STK::IMixtureComposer::randomTik(), STK::MixtureSemiLearner::randomTik(), STK::IMixtureComposer::randomZi(), STK::MixtureSemiLearner::randomZi(), STK::IMixtureLearner::setClassLabels(), STK::IMixtureLearner::setClassLabels(), setMixtureParameters(), setMixtureParameters(), and setProportions().

tik_

CArrayXX STK::IMixtureStatModel::tik_

protected

The tik probabilities.

Definition at line 333 of file STK_IMixtureStatModel.h.

Referenced by computeICL(), STK::IMixtureComposer::cStep(), STK::MixtureSemiLearner::cStep(), STK::IMixtureComposer::cStep(), STK::IMixtureComposer::eStep(), STK::MixtureSemiLearner::eStep(), STK::IMixtureComposer::eStep(), STK::IMixtureComposer::initializeMixtureParameters(), STK::MixtureSemiLearner::initializeMixtureParameters(), STK::IMixtureComposer::initializeStep(), STK::IMixtureLearner::mapStep(), STK::IMixtureComposer::mapStep(), STK::IMixtureComposer::pStep(), STK::IMixtureComposer::randomTik(), STK::MixtureSemiLearner::randomTik(), STK::IMixtureLearner::setClassLabels(), STK::IMixtureLearner::setClassLabels(), STK::MixtureSemiLearner::setLabels(), setMixtureParameters(), setMixtureParameters(), STK::IMixtureComposer::sStep(), and tik().

tk_

CPointX STK::IMixtureStatModel::tk_

protected

The sum of the columns of tik_.

Definition at line 335 of file STK_IMixtureStatModel.h.

Referenced by STK::IMixtureComposer::cStep(), STK::MixtureSemiLearner::cStep(), STK::IMixtureComposer::eStep(), STK::MixtureSemiLearner::eStep(), STK::IMixtureComposer::eStep(), STK::IMixtureComposer::initializeMixtureParameters(), STK::IMixtureComposer::initializeStep(), STK::IMixtureComposer::randomTik(), STK::MixtureSemiLearner::randomTik(), STK::IMixtureLearner::setClassLabels(), STK::IMixtureLearner::setClassLabels(), setMixtureParameters(), setMixtureParameters(), and tk().

v_mixtures_

std::vector<IMixture*> STK::IMixtureStatModel::v_mixtures_

protected

vector of pointers to the mixtures components

Definition at line 340 of file STK_IMixtureStatModel.h.

Referenced by computeNbFreeParameters(), computeNbMissingValues(), STK::MixtureComposer::create(), STK::MixtureComposerFixedProp::create(), STK::MixtureLearner::create(), STK::MixtureComposer::createComposer(), STK::MixtureLearner::createLearner(), STK::MixtureComposer::finalizeStep(), STK::MixtureLearner::finalizeStep(), getMixture(), IMixtureStatModel(), STK::MixtureComposer::imputationStep(), STK::MixtureLearner::imputationStep(), initializeStep(), STK::MixtureComposer::lnComponentProbability(), STK::MixtureLearner::lnComponentProbability(), STK::MixtureComposer::paramUpdateStep(), STK::MixtureLearner::paramUpdateStep(), STK::MixtureComposer::randomInit(), STK::MixtureLearner::randomInit(), registerMixture(), STK::MixtureComposer::releaseIntermediateResults(), STK::MixtureLearner::releaseIntermediateResults(), releaseMixture(), STK::MixtureComposer::samplingStep(), STK::MixtureLearner::samplingStep(), STK::MixtureComposer::setParametersStep(), STK::MixtureLearner::setParametersStep(), STK::MixtureComposer::storeIntermediateResults(), STK::MixtureLearner::storeIntermediateResults(), v_mixtures(), STK::MixtureComposer::writeParameters(), STK::MixtureLearner::writeParameters(), and ~IMixtureStatModel().

zi_

CVectorXi STK::IMixtureStatModel::zi_

protected

The zi class label.

Definition at line 337 of file STK_IMixtureStatModel.h.

Referenced by STK::IMixtureComposer::cStep(), STK::MixtureSemiLearner::eStep(), STK::IMixtureComposer::eStep(), STK::MixtureSemiLearner::initializeMixtureParameters(), STK::IMixtureComposer::mapStep(), STK::IMixtureComposer::mapStep(), STK::MixtureSemiLearner::MixtureSemiLearner(), STK::IMixtureComposer::randomZi(), STK::MixtureSemiLearner::randomZi(), STK::IMixtureLearner::setClassLabels(), STK::IMixtureLearner::setClassLabels(), STK::MixtureSemiLearner::setLabels(), setMixtureParameters(), setMixtureParameters(), STK::IMixtureComposer::sStep(), STK::IMixtureComposer::sStep(), and zi().

---

The documentation for this class was generated from the following files:

• STK_IMixtureStatModel.h
• STK_IMixtureStatModel.cpp