libbpp-core-dev 2.1.0-1 / usr / include / Bpp / Numeric / Prob / DirichletDiscreteDistribution.h

//
// File: DirichletDiscreteDistribution.h
// Created by: Laurent Guéguen
// Created on: jeudi 2 septembre 2010, à 17h 03
//

/*
   Copyright or © or Copr. Bio++ Development Team, (November 19, 2004)

   This software is a computer program whose purpose is to provide classes
   for numerical calculus.

   This software is governed by the CeCILL  license under French law and
   abiding by the rules of distribution of free software.  You can  use,
   modify and/ or redistribute the software under the terms of the CeCILL
   license as circulated by CEA, CNRS and INRIA at the following URL
   "http://www.cecill.info".

   As a counterpart to the access to the source code and  rights to copy,
   modify and redistribute granted by the license, users are provided only
   with a limited warranty  and the software's author,  the holder of the
   economic rights,  and the successive licensors  have only  limited
   liability.

   In this respect, the user's attention is drawn to the risks associated
   with loading,  using,  modifying and/or developing or reproducing the
   software by the user in light of its specific status of free software,
   that may mean  that it is complicated to manipulate,  and  that  also
   therefore means  that it is reserved for developers  and  experienced
   professionals having in-depth computer knowledge. Users are therefore
   encouraged to load and test the software's suitability as regards their
   requirements in conditions enabling the security of their systems and/or
   data to be ensured and,  more generally, to use and operate it in the
   same conditions as regards security.

   The fact that you are presently reading this means that you have had
   knowledge of the CeCILL license and that you accept its terms.
 */

#ifndef _DIRICHLETDISCRETEDISTRIBUTION_H_
#define _DIRICHLETDISCRETEDISTRIBUTION_H_

#include "MultipleDiscreteDistribution.h"
#include "BetaDiscreteDistribution.h"
#include "../VectorTools.h"
#include "../ParameterAliasable.h"
#include "../../Exceptions.h"
#include "../../Io/OutputStream.h"

namespace bpp
{
/**
 * @brief Discretized Dirichlet distribution. If the distribution is
 * in n dimensions, the domain is the n-1 simplex.
 *
 * The discretization is made in the order of these dimensions. Then
 * the order of the dimensions is not neutral on the resulting
 * categories.
 *
 * Here is an example of the discretization process: if the
 * distribution is in 3 dimensions, and the number of categories is
 * 4 in the first dimension, 3 in the second and 5 in the third, the
 * first interval is split in 4 equi-probable intervals, and each
 * interval is split on the second dimension in 3 equi-probable
 * intervals, and each of the 12 resulting 2D intervals is split in
 * 5 equi-probable 3D-intervals. Eventually, there are 60
 * equi-probable categories.
 *
 * In the same example, if the distribution is followed by the
 * random vector @f$ (X_1,X_2,X_3) @f$, the 1D distribution used for
 * @f$ X_1 @f$ is the marginal distribution, which discretized
 * values are say, @f$ v_1^1, v_1^2, v_1^3, v_1^4 @f$, then the four
 * 1D distributions used for @f$ X_2 @f$ are the conditional
 * marginal distributions @f$ (X_2|X_1=v_1^1), (X_2|X_1=v_1^2),
 * (X_2|X_1=v_1^3), (X_2|X_1=v_1^4) @f$. And in a similar way for
 * @f$ X_3 @f$, with the 12 resulting conditions on @f$ X_1 @f$ and
 * @f$ X_2 @f$.
 *
 *
 * It uses the AbstractDiscreteDistribution comparator class to deal
 * with double precision. By default, category values that differ
 * less than 10E-9 will be considered identical.
 *
 *
 * The successive discretization process is done using the
 * distributions, if the parameters are @f$ (\alpha_1, ...,
 * \alpha_n)@f$:
 *
 * @f$ X_1 @f$ follows Beta(@f$ \alpha_1, \sum_{i=2}^n \alpha_i @f$).
 *
 * @f$ X_j @f$ follows @f$ (1-\sum_{i=1}^{j-1} X_i) @f$ *
 * Beta(@f$ \alpha_j, \sum_{i=j+1}^n \alpha_i @f$).
 *
 * @f$ X_n = 1 - \sum_{i=1}^{n-1} X_i @f$.
 *
 *
 * The parameters are: alpha_1, ... alpha_n @f$ \in [0.0001;\infty[
 * @f$.
 */

class DirichletDiscreteDistribution :
  public MultipleDiscreteDistribution,
  public AbstractParameterAliasable
{
private:
  std::vector<BetaDiscreteDistribution* > vpBDD_;

public:
  /**
   * @brief Build a new discretized Dirichlet distribution
   *
   * @param vn the vector of the dim-1 numbers of categories to use.
   * @param valpha the vector of the dim alpha parameters.
   */
  DirichletDiscreteDistribution(std::vector<size_t> vn, Vdouble valpha);

  ~DirichletDiscreteDistribution();

  DirichletDiscreteDistribution* clone() const
  {
    return new DirichletDiscreteDistribution(*this);
  }

protected:
  void applyParameters();

public:
  std::string getName() const {return("Dirichlet");}
  
  void fireParameterChanged(const ParameterList& parameters);

  /**
   * @return The number of categories
   */
  size_t getNumberOfCategories() const;

  /**
   * @param Vvalue
   * @return The vector of categoryIndex of the classes the value is
   * in. Throws a ConstraintException if the value is off the domain
   * of the DirichletDiscreteDistribution.
   */
  Vdouble getValueCategory(Vdouble& Vvalue) const;

  VVdouble getCategories() const;

  /**
   * @param category The vector of values associated to the class.
   * @return The probability associated to a given class.
   */
  virtual double getProbability(Vdouble& category) const;

  /**
   * @brief Draw a random vector from this distribution.
   *
   * This vector will be one of the class values, drawn according
   * to the class probabilities.
   *
   * @return A random number according to this distribution.
   */

  Vdouble rand() const;

  /**
   * @brief Draw a random vector from the continuous version of this
   * distribution.
   *
   * @return A random vector according to this distribution.
   */
  Vdouble randC() const;

protected:
  void discretize(Vdouble& valpha);
};
} // end of namespace bpp.

#endif  // _DIRICHLETDISCRETEDISTRIBUTION_H_