STK::DiagGaussianBase< Derived > Class Template Reference

Base class for the diagonal Gaussian models. More...

#include <STK_DiagGaussianBase.h>

Inheritance diagram for STK::DiagGaussianBase< Derived >:

Public Types
typedef IMixtureDensity< Derived >	Base
Public Types inherited from STK::IMixtureDensity< Derived >
typedef hidden::MixtureTraits< Derived >::Array	Array
typedef hidden::MixtureTraits< Derived >::Parameters	Parameters
typedef hidden::Traits< Array >::Type	Type

Public Member Functions
Real const &	mean (int k, int j) const
Real const &	sigma (int k, int j) const
void	initializeModelImpl ()
	Initialize the parameters of the model.
Real	lnComponentProbability (int i, int k) const
template<class Weights >
Real	impute (int i, int j, Weights const &pk) const
Real	rand (int i, int j, int k) const
template<class Array >
void	getParameters (Array &params) const
	This function is used in order to get the current values of the means and standard deviations.
void	writeParameters (CArrayXX const *p_tik, ostream &os) const
	This function can be used to write summary of parameters to the output stream.
int	nbCluster () const
int	nbSample () const
Array const *const &	p_data () const
Public Member Functions inherited from STK::IMixtureDensity< Derived >
	~IMixtureDensity ()
	destructor
int	nbCluster () const
int	nbSample () const
Real	lnNbSample () const
Array const *const &	p_data () const
Parameters const &	param () const
void	setData (Array const &data)
	Set the data set.
void	setData (Array const &data, int nbRow, int nbCol, bool byRow=true)
	Set the data set and give dimensions.
bool	initializeStep ()
	This function will be called at the beginning of the estimation process once the model is created and data is set.
void	setParametersStep ()
	set the parameters obtained with the intermediate results and release the intermediate results.
void	finalizeStep ()
	This function will be called once the model is estimated.
template<class Weights >
Type	sample (int i, int j, Weights const &tk) const
Public Member Functions inherited from STK::IRecursiveTemplate< Derived >
Derived &	asDerived ()
	static cast : return a reference of this with a cast to the derived class.
Derived const &	asDerived () const
	static cast : return a const reference of this with a cast to the derived class.
Derived *	asPtrDerived ()
	static cast : return a ptr on a Derived of this with a cast to the derived class.
Derived const *	asPtrDerived () const
	static cast : return a ptr on a constant Derived of this with a cast to the derived class.
Derived *	clone () const
	create a leaf using the copy constructor of the Derived class.
Derived *	clone (bool isRef) const
	create a leaf using the copy constructor of the Derived class and a flag determining if the clone is a reference or not.

Public Attributes
Parameters	param_
	parameters of the derived mixture model.

Protected Member Functions
	DiagGaussianBase (int nbCluster)
	default constructor
	DiagGaussianBase (DiagGaussianBase const &model)
	copy constructor
	~DiagGaussianBase ()
	destructor
void	randomMean (CArrayXX const *p_tik)
	sample randomly the mean of each component by sampling randomly a row of the data set.
bool	updateMean (CArrayXX const *p_tik)
	compute the weighted mean of a Gaussian mixture.
Protected Member Functions inherited from STK::IMixtureDensity< Derived >
	IMixtureDensity (int nbCluster)
	Default constructor.
	IMixtureDensity (IMixtureDensity const &model)
	copy constructor.
void	initializeModel ()
	Initialize the model before its first use.
bool	initializeStepImpl ()
	default implementation of initializeStepImpl (do nothing and return true)
void	finalizeStepImpl ()
	default implementation of finalizeStepImpl (do nothing)
void	setNbSample (int nbSample)
	Set the number of sample of the model (needed by kernel models)
Protected Member Functions inherited from STK::IRecursiveTemplate< Derived >
	IRecursiveTemplate ()
	constructor.
	~IRecursiveTemplate ()
	destructor.

Additional Inherited Members
Protected Attributes inherited from STK::IMixtureDensity< Derived >
Parameters	param_
	parameters of the derived mixture model.

Detailed Description

template<class Derived>
class STK::DiagGaussianBase< Derived >

Base class for the diagonal Gaussian models.

Definition at line 53 of file STK_DiagGaussianBase.h.

Member Typedef Documentation

Base

template<class Derived >

typedef IMixtureDensity<Derived > STK::DiagGaussianBase< Derived >::Base

Definition at line 56 of file STK_DiagGaussianBase.h.

Constructor & Destructor Documentation

DiagGaussianBase() [1/2]

template<class Derived >

STK::DiagGaussianBase< Derived >::DiagGaussianBase ( int  nbCluster )

inlineprotected

default constructor

Parameters

nbCluster

number of cluster in the model

Definition at line 66 of file STK_DiagGaussianBase.h.

: Base(nbCluster) {}

DiagGaussianBase() [2/2]

template<class Derived >

STK::DiagGaussianBase< Derived >::DiagGaussianBase ( DiagGaussianBase< Derived > const &  model )

inlineprotected

copy constructor

Parameters

model

The model to copy

Definition at line 70 of file STK_DiagGaussianBase.h.

: Base(model) {}

~DiagGaussianBase()

template<class Derived >

STK::DiagGaussianBase< Derived >::~DiagGaussianBase ( )

inlineprotected

destructor

Definition at line 72 of file STK_DiagGaussianBase.h.

{}

Member Function Documentation

getParameters()

template<class Derived >

template<class Array >

void STK::DiagGaussianBase< Derived >::getParameters

(

Array &

params

)

const

This function is used in order to get the current values of the means and standard deviations.

Parameters

[out]

params

the array with the parameters of the mixture.

Definition at line 187 of file STK_DiagGaussianBase.h.

{
  int nbClust = nbCluster();
  params.resize(2*nbClust, p_data()->cols());
  for (int k= 0; k < nbClust; ++k)
  {
    for (int j= params.beginCols();  j< params.endCols(); ++j)
    {
      params(baseIdx+2*k  , j) = mean(baseIdx + k, j);
      params(baseIdx+2*k+1, j) = sigma(baseIdx + k, j);
    }
  }
}

References STK::baseIdx, and STK::mean().

impute()

template<class Derived >

template<class Weights >

Real STK::DiagGaussianBase< Derived >::impute	(	int	i,
		int	j,
		Weights const &	pk
	)		const

Returns

an imputation value for the jth variable of the ith sample

Parameters

i,j	indexes of the data to impute
pk	the probabilities of each class for the ith individual

Definition at line 140 of file STK_DiagGaussianBase.h.

{
  Real sum = 0.;
  for (int k= pk.begin(); k < pk.end(); ++k)
  { sum += pk[k] * mean(k,j);}
  return sum;
}

References STK::mean(), and STK::sum().

initializeModelImpl()

template<class Derived >

void STK::DiagGaussianBase< Derived >::initializeModelImpl ( )

inline

Initialize the parameters of the model.

Definition at line 81 of file STK_DiagGaussianBase.h.

{ param_.resize(p_data()->cols());}

References STK::DiagGaussianBase< Derived >::p_data(), and STK::DiagGaussianBase< Derived >::param_.

lnComponentProbability()

template<class Derived >

Real STK::DiagGaussianBase< Derived >::lnComponentProbability	(	int	i,
		int	k
	)		const

Returns

value of the log-probability of the i-th sample in the k-th cluster.

Parameters

i,k	indexes of the sample and of the component

Definition at line 123 of file STK_DiagGaussianBase.h.

{
  Real sum =0.;
  for (int j=p_data()->beginCols(); j<p_data()->endCols(); ++j)
  {
    if (sigma(k,j))
    { sum += Law::Normal::lpdf(p_data()->elt(i,j), mean(k, j), sigma(k,j));}
  }
  return sum;
}

References STK::Law::Normal::lpdf(), STK::mean(), and STK::sum().

mean()

template<class Derived >

Real const & STK::DiagGaussianBase< Derived >::mean ( int  k, int  j  ) const

inline

Returns

mean of the kth cluster and jth variable

Definition at line 76 of file STK_DiagGaussianBase.h.

{ return param_.mean(k,j);}

References STK::DiagGaussianBase< Derived >::param_.

Referenced by STK::DiagGaussianBase< Derived >::rand().

nbCluster()

template<class Derived >

int STK::IMixtureDensity< Derived >::nbCluster ( ) const

inline

Returns

the number of cluster

Definition at line 124 of file STK_IMixtureDensity.h.

{ return nbCluster_;}

nbSample()

template<class Derived >

int STK::IMixtureDensity< Derived >::nbSample ( ) const

inline

Returns

the total available observations

Definition at line 126 of file STK_IMixtureDensity.h.

{ return nbSample_;}

p_data()

template<class Derived >

Array const *const & STK::IMixtureDensity< Derived >::p_data ( ) const

inline

Returns

a pointer on the current data set

Definition at line 131 of file STK_IMixtureDensity.h.

{ return p_dataij_;}

Referenced by STK::DiagGaussianBase< Derived >::initializeModelImpl().

rand()

template<class Derived >

Real STK::DiagGaussianBase< Derived >::rand ( int  i, int  j, int  k  ) const

inline

Returns

a simulated value for the jth variable of the ith sample in the kth cluster

Parameters

i,j,k

indexes of the data to simulate

Definition at line 95 of file STK_DiagGaussianBase.h.

    { return Law::Normal::rand(mean(k, j), sigma(k,j));}

References STK::DiagGaussianBase< Derived >::mean(), STK::Law::Normal::rand(), and STK::DiagGaussianBase< Derived >::sigma().

randomMean()

template<class Derived >

void STK::DiagGaussianBase< Derived >::randomMean ( CArrayXX const *  p_tik )

protected

sample randomly the mean of each component by sampling randomly a row of the data set.

Definition at line 149 of file STK_DiagGaussianBase.h.

{
  // indexes array
  VectorXi indexes(p_data()->rows());
  for(int i=indexes.begin(); i< indexes.end(); ++i) { indexes[i] = i;}
  Range rind(p_data()->rows());
  // sample without repetition
  for (int k= p_tik->beginCols(); k < p_tik->endCols(); ++k)
  {
    // random number in [0, end-k[
    int i = Law::UniformDiscrete::rand(rind.begin(), rind.end()-1);
    // get ith individuals
    param_.mean_[k] = p_data()->row(indexes[i]);
    // exchange it with nth
    indexes.swap(i, rind.lastIdx());
    // decrease
    rind.decLast(1);
  }
}

References STK::Law::UniformDiscrete::rand().

sigma()

template<class Derived >

Real const & STK::DiagGaussianBase< Derived >::sigma ( int  k, int  j  ) const

inline

Returns

sigma of the kth cluster and jth variable

Definition at line 78 of file STK_DiagGaussianBase.h.

{ return param_.sigma(k,j);}

References STK::DiagGaussianBase< Derived >::param_.

Referenced by STK::DiagGaussianBase< Derived >::rand().

updateMean()

template<class Derived >

bool STK::DiagGaussianBase< Derived >::updateMean ( CArrayXX const *  p_tik )

protected

compute the weighted mean of a Gaussian mixture.

Definition at line 171 of file STK_DiagGaussianBase.h.

{
  for (int k= p_tik->beginCols(); k < p_tik->endCols(); ++k)
  {
    for (int j=p_data()->beginCols(); j< p_data()->endCols(); ++j)
    { param_.mean_[k][j] = p_data()->col(j).wmean(p_tik->col(k));}
  }
  return true;
}

References STK::ICArray< Derived >::col().

writeParameters()

template<class Derived >

void STK::DiagGaussianBase< Derived >::writeParameters	(	CArrayXX const *	p_tik,
		ostream &	os
	)		const

This function can be used to write summary of parameters to the output stream.

Parameters

p_tik	a constant pointer on the posterior probabilities
os	Stream where you want to write the summary of parameters.

Definition at line 207 of file STK_DiagGaussianBase.h.

{
  CPointX m(p_data()->cols()), s(p_data()->cols());
  for (int k= p_tik->beginCols(); k < p_tik->endCols(); ++k)
  {
    // store sigma values in an array for a nice output
    for (int j= s.begin();  j < s.end(); ++j)
    { m[j] = mean(k,j); s[j] = sigma(k,j);}
    os << _T("---> Component ") << k << _T("\n");
    os << _T("mean = ") << m;
    os << _T("sigma = ")<< s;
  }
}

References _T, and STK::mean().

Member Data Documentation

param_

template<class Derived >

Parameters STK::IMixtureDensity< Derived >::param_

parameters of the derived mixture model.

Should be an instance of the STK::ModelParameters struct.

See also

STK::ModelParameters

Definition at line 180 of file STK_IMixtureDensity.h.

Referenced by STK::DiagGaussianBase< Derived >::initializeModelImpl(), STK::DiagGaussianBase< Derived >::mean(), and STK::DiagGaussianBase< Derived >::sigma().

---

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

• STK_DiagGaussianBase.h