source: trunk/yat/classifier/NBC.h @ 1437

#ifndef _theplu_yat_classifier_nbc_ 
#define _theplu_yat_classifier_nbc_ 

// $Id: NBC.h 1437 2008-08-25 17:55:00Z peter $

/*
  Copyright (C) 2006 Jari Häkkinen, Peter Johansson, Markus Ringnér
  Copyright (C) 2007 Jari Häkkinen, Peter Johansson
  Copyright (C) 2008 Peter Johansson, Markus Ringnér

  This file is part of the yat library, http://dev.thep.lu.se/yat

  The yat library is free software; you can redistribute it and/or
  modify it under the terms of the GNU General Public License as
  published by the Free Software Foundation; either version 2 of the
  License, or (at your option) any later version.

  The yat 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
  General Public License for more details.

  You should have received a copy of the GNU 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.
*/

#include "SupervisedClassifier.h"
#include "yat/utility/Matrix.h"

namespace theplu {
namespace yat {
namespace classifier {  

  class MatrixLookup;
  class MatrixLookupWeighted;
  class Target;

  /**
     @brief Naive Bayesian Classifier.
  
     Each class is modelled as a multinormal distribution with
     features being independent: \f$ P(x|c) \propto \prod
     \frac{1}{\sqrt{2\pi\sigma_i^2}} \exp \left(
     -\frac{(x_i-\mu_i)^2}{2\sigma_i^2)} \right)\f$
  */
  class NBC : public SupervisedClassifier
  {
  
  public:
    ///
    /// @brief Constructor 
    ///
    NBC(void);
    

    ///
    /// @brief Destructor
    ///
    virtual ~NBC();


    NBC* make_classifier(void) const;
    
    ///
    /// \brief Train the NBC using training data and targets.
    ///
    /// For each class mean and variance are estimated for each
    /// feature (see statistics::Averager for details).
    ///
    /// If there is only one (or zero) samples in a class, parameters
    /// cannot be estimated. In that case, parameters are set to NaN
    /// for that particular class.
    ///
    void train(const MatrixLookup&, const Target&);

    ///
    /// \brief Train the NBC using weighted training data and
    /// targets.
    ///
    /// For each class mean and variance are estimated for each
    /// feature (see statistics::AveragerWeighted for details).
    ///
    /// To estimate the parameters of a class, each feature of the
    /// class must have at least two non-zero data points. Otherwise
    /// the parameters are set to NaN and any prediction will result
    /// in NaN for that particular class.
    ///
    void train(const MatrixLookupWeighted&, const Target&);
    
    /**
       \brief Predict samples using unweighted data

       Each sample (column) in \a data is predicted and predictions
       are returned in the corresponding column in passed \a
       result. Each row in \a result corresponds to a class. The
       prediction is the estimated probability that sample belong to
       class \f$ j \f$:

       \f$ P_j = \frac{1}{Z}\prod_i\frac{1}{\sqrt{2\pi\sigma_i^2}}
       \exp\left(-\frac{(x_i-\mu_i)^2}{2\sigma_i^2}\right)\f$, where \f$ \mu_i
       \f$ and \f$ \sigma_i^2 \f$ are the estimated mean and variance,
       respectively. Z is chosen such that total probability equals unity, \f$
       \sum P_j = 1 \f$.

       \note If parameters could not be estimated during training, due
       to lack of number of sufficient data points, the output for
       that class is NaN and not included in calculation of
       normalization factor \f$ Z \f$.
    */
    void predict(const MatrixLookup& data, utility::Matrix& result) const;

    /**
       \brief Predict samples using weighted data

       Each sample (column) in \a data is predicted and predictions
       are returned in the corresponding column in passed \a
       result. Each row in \a result corresponds to a class. The
       prediction is the estimated probability that sample belong to
       class \f$ j \f$:

       \f$ P_j = \frac{1}{Z} \exp\left(-N\frac{\sum
       {w_i(x_i-\mu_i)^2}/(2\sigma_i^2)}{\sum w_i}\right)
       \prod_i\frac{1}{\sqrt{2\pi\sigma_i^2}}\f$, where \f$ \mu_i \f$
       and \f$ \sigma_i^2 \f$ are the estimated mean and variance,
       respectively. Z is chosen such that total probability equals
       unity, \f$ \sum P_j = 1 \f$.

       \note If parameters could not be estimated during training, due
       to lack of number of sufficient data points, the output for
       that class is NaN and not included in calculation of
       normalization factor \f$ Z \f$.
     */
    void predict(const MatrixLookupWeighted& data,utility::Matrix& result) const;


  private:
    void standardize_lnP(utility::Matrix& prediction) const;

    utility::Matrix centroids_;
    utility::Matrix sigma2_;

    double sum_logsigma(size_t i) const;


  };
  
}}} // of namespace classifier, yat, and theplu

#endif