STK_Gamma_ak_bj.h

Go to the documentation of this file.

/*--------------------------------------------------------------------*/
/*     Copyright (C) 2004-2016  Serge Iovleff, Université Lille 1, Inria
 
    This program 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 of the
    License, or (at your option) any later version.
 
    This program 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 program; if not, write to the
    Free Software Foundation, Inc.,
    59 Temple Place,
    Suite 330,
    Boston, MA 02111-1307
    USA
 
    Contact : S..._DOT_I..._AT_stkpp.org (see copyright for ...)
*/
 
/*
 * Project: stkpp::Clustering
 * created on: 5 sept. 2013
 * Author:  iovleff, serge.iovleff@stkpp.org
 **/
 
#ifndef STK_GAMMA_AK_BJ_H
#define STK_GAMMA_AK_BJ_H
 
#include <STatistiK/include/STK_Law_Exponential.h>
#include "../GammaModels/STK_GammaBase.h"
 
#define MAXITER 400
#define TOL 1e-8
 
namespace STK
{
template<class Array>class Gamma_ak_bj;
 
namespace hidden
{
template<class Array_>
struct MixtureTraits< Gamma_ak_bj<Array_> >
{
  typedef Array_ Array;
  typedef ModelParameters<Clust::Gamma_ak_bj_> Parameters;
};
 
} // namespace Clust
 
 
template<class Array>
class Gamma_ak_bj: public GammaBase< Gamma_ak_bj<Array> >
{
  public:
    typedef GammaBase<Gamma_ak_bj<Array> > Base;
    using Base::param_;
    using Base::p_data;
    using Base::meanjk;
    using Base::variancejk;
 
    Gamma_ak_bj( int nbCluster): Base(nbCluster) {}
    Gamma_ak_bj( Gamma_ak_bj const& model): Base(model) {}
    ~Gamma_ak_bj() {}
    void randomInit( CArrayXX const* const& p_tik, CPointX const* const& p_tk) ;
    bool run( CArrayXX const* const& p_tik, CPointX const* const& p_tk) ;
    inline int computeNbFreeParameters() const
    { return this->nbCluster()+ p_data()->sizeCols();}
};
 
/* Initialize randomly the parameters of the Gaussian mixture. The centers
 *  will be selected randomly among the data set and the standard-deviation
 *  will be set to 1.
 */
template<class Array>
void Gamma_ak_bj<Array>::randomInit( CArrayXX const* const& p_tik, CPointX const* const& p_tk) 
{
    // compute moments
    this->moments(p_tik);
  // simulates ak
  for (int k= p_tik->beginCols(); k < p_tik->endCols(); ++k)
  {
    Real value= 0.;
    for (int j=p_data()->beginCols(); j < p_data()->endCols(); ++j)
    {
      Real mean = meanjk(j,k), variance = variancejk(j,k);
      value += mean*mean/variance;
    }
    param_.shape_[k]= Law::Exponential::rand(value/(p_data()->sizeCols()));
  }
  // simulate bj
  for (int j=p_data()->beginCols(); j < p_data()->endCols(); ++j)
  {
    Real value= 0.;
    for (int k= p_tik->beginCols(); k < p_tik->endCols(); ++k)
    {
      Real mean = meanjk(j,k), variance = variancejk(j,k);
      value += p_tk->elt(k) * variance/mean;
    }
    param_.scale_[j] = Law::Exponential::rand(value/(this->nbSample()));
  }
#ifdef STK_MIXTURE_VERY_VERBOSE
  stk_cout << _T(" Gamma_ak_bj<Array>::randomInit done\n");
#endif
}
 
/* Compute the weighted mean and the common variance. */
template<class Array>
bool Gamma_ak_bj<Array>::run( CArrayXX const* const& p_tik, CPointX const* const& p_tk) 
{
  if (!this->moments(p_tik)) { return false;}
  // start estimations of the ajk and bj
  Real qvalue = this->qValue(p_tik, p_tk);
  int iter;
  for(iter=0; iter<MAXITER; ++iter)
  {
    // compute ak
    for (int k= p_tik->beginCols(); k < p_tik->endCols(); ++k)
    {
      // moment estimate and oldest value
      Real x0 = (param_.mean_[k].square()/param_.variance_[k]).mean();
      Real x1 =  param_.shape_[k];
      if ((x0 <=0.) || !Arithmetic<Real>::isFinite(x0)) return false;
 
      // compute shape
      hidden::invPsi f((param_.meanLog_[k] - param_.scale_.log()).mean());
      Real a =  Algo::findZero(f, x0, x1, TOL);
 
      if (!Arithmetic<Real>::isFinite(a))
      {
        param_.shape_[k]= x0; // use moment estimate
#ifdef STK_MIXTURE_DEBUG
        stk_cout << _T("ML estimation failed in Gamma_ak_bj::run( CArrayXX const* const& p_tik, CPointX const* const& p_tk) \n");
        stk_cout << "x0 =" << x0 << _T("\n";);
        stk_cout << "f(x0) =" << f(x0) << _T("\n";);
        stk_cout << "x1 =" << x1 << _T("\n";);
        stk_cout << "f(x1) =" << f(x1) << _T("\n";);
#endif
      }
      else { param_.shape_[k]= a;}
    }
    // update all the b^j
    for (int j=p_data()->beginCols(); j<p_data()->endCols(); ++j)
    {
      Real num = 0., den = 0.;
      for (int k= p_tik->beginCols(); k < p_tik->endCols(); ++k)
      {
        num += param_.mean_[k][j] * p_tk->elt(k);
        den += param_.shape_[k]   * p_tk->elt(k);
      }
      // compute b_j
      Real b = num/den;
      // divergence
      if (!Arithmetic<Real>::isFinite(b)) { return false;}
      param_.scale_[j] = b;
    }
    // check convergence
    Real value = this->qValue(p_tik, p_tk);
#ifdef STK_MIXTURE_DEBUG
    if (value < qvalue)
    {
      stk_cout << _T("In Gamma_ak_bj::run( CArrayXX const* const& p_tik, CPointX const* const& p_tk) : run( CArrayXX const* const& p_tik, CPointX const* const& p_tk)  diverge\n");
      stk_cout << _T("New value =") << value << _T(", qvalue =") << qvalue << _T("\n");
    }
#endif
    if ((value - qvalue) < TOL) break;
    qvalue = value;
  }
#ifdef STK_MIXTURE_DEBUG
  if (iter == MAXITER)
  {
    stk_cout << _T("In Gamma_ak_bj::run( CArrayXX const* const& p_tik, CPointX const* const& p_tk) : run( CArrayXX const* const& p_tik, CPointX const* const& p_tk)  did not converge\n");
    stk_cout << _T("qvalue =") << qvalue << _T("\n");
  }
#endif
  return true;
}
 
 
}  // namespace STK
 
#undef MAXITER
#undef TOL
 
#endif /* STK_Gamma_AK_BJ_H */